Epidemiology of invasive fungal infections in the intensive care unit: results of a multicenter Italian survey (AURORA Project)

Purpose The aims of this study are to evaluate the epidemiology of invasive fungal infections (IFIs) in patients admitted to an intensive care unit (ICU) in Southern Italy and the in vitro antifungal susceptibility of isolates. Methods A surveillance program was implemented in 18 ICUs. IFI cases were recorded using a standardized form. Results A total of 105 episodes of IFIs occurred in 5,561 patients during the 18-month study. The main infections were caused by yeasts, more than filamentous fungi (overall incidence of 16.5 cases per 1,000 admissions and 2.3 cases per 1,000 admissions, respectively). The overall crude mortality rate was high (42.8 %), particularly for mold infections (61.5 %). All yeast infections were Candida bloodstream infections. Over half (59.8 %) were caused by Candida non-albicans, with C. parapsilosis being the most common (61.8 %). In the multivariate model, trauma admission diagnosis, prolonged stay in the ICU, and parenteral nutrition were independently associated with candidemia due to C. parapsilosis [odds ratio (OR) 3.5, (1.8–5.2); OR 3.5, (1.02–3.5); OR 3.6, (1.28–6.99), respectively]. Among mold infections, 12 patients suffered from invasive pulmonary aspergillosis, with Aspergillus fumigatus as the predominant pathogen (41.7 %). One case of brain scedosporiosis was identified. Overall, azoles and echinocandins resistance was uncommon. Conclusions Candida non-albicans species are the most frequent cause of candidemia in ICU patients. Mold infections are associated with a high mortality rate. This study confirms the importance of the epidemiological surveillance on IFIs in the ICU setting for documenting species distribution and antimicrobial susceptibility patterns to guide therapeutic choices.


Introduction
The incidence of invasive fungal infections (IFIs) has increased significantly worldwide, representing an important infective complication in hospitalized patients. In particular, critically ill patients are highly susceptible to IFIs: these diseases are very worrisome in the intensive care unit (ICU) due to the complexity of the patients' underlying conditions.
The epidemiology of IFIs is generally characterized by geographical and temporal variability. Different incidence rates and new emerging species have been revealed during the last 20 years: the European data estimate an incidence of Candida bloodstream infection (BSI) ranging between 6.7 and 54 per 1,000 ICU admissions, with a mortality of 33.9-61.8 % [1,2]. Until recently, Candida albicans was the most prominent species in ICU patients [3]. However, a shift towards Candida non-albicans (CnA) throughout the world has been reported: C. glabrata accounts for around 20 % of ICU-associated BSI in some areas [1,4,5], while C. parapsilosis is the most relevant non-albicans species elsewhere [6][7][8].
Due to the poor outcome related to IFIs in critically ill patients, the knowledge of local epidemiologic trends and antifungal susceptibility of etiological agents is critical. We aimed to determine the contemporary epidemiology, management, and outcome of IFIs in ICU patients in Southern Italy, as well as the in vitro antifungal susceptibility of isolates.

Design of the study
This study is a subset of the AURORA Project [12,13], a multicenter, observational study, performed between February 2007 and August 2008.
A surveillance program was implemented in 18 ICUs of 16 hospitals. All consecutive adult patients (C18 years old) who had a documented IFI, either on admission or during stay, were enrolled. The Institutional Review Board approved the protocol. Informed consent was obtained from patients or their representative.
In order to standardize clinical recruitment and microbiological procedures, a training course was performed before starting the study, and a manual was distributed to all participants. In addition, a memento was carried out monthly by e-mail or by phone, to both microbiologists and clinicians, to estimate the fullness of reporting and to ensure that all cases were notified. According to protocol, each hospital-associated microbiology laboratory carried out a mycological analysis on biological samples from potentially infected sites, and all fungal species were isolated, identified, and stored at -80°C. The galactomannan (GM) antigen test was performed in patients with clinical signs and/or symptoms suggestive of invasive aspergillosis (e.g., development of pulmonary infiltrates on chest X-ray, fever refractory to at least 3 days of appropriate antibiotics, pleuritic chest pain).
For each IFI case, the participating center had to complete an electronic report form including demographic and clinical data [age, gender, comorbidities, ICU admission typology, severity of illness on ICU admission, organ dysfunctions on enrollment, length of stay, presence of central venous catheter (CVC) and its removal or not on IFI diagnosis, steroid therapy, total parenteral nutrition, mechanical ventilation, concomitant bacterial infections and antibiotic therapy, antifungal prophylaxis and treatment], microbiological data (e.g., date of first positive culture, etiological agents, fungal colonization), radiological and/or histological findings, and outcome. In order to confirm the identification of fungal isolates, test their in vitro antifungal susceptibility, and analyze the data, both the isolates and the electronic reports were sent to the Coordinating Centre (Laboratory of Mycology-Department of Biomedical Science and Human Oncology, University of Bari Aldo Moro, Bari, Italy).

Definitions
Yeast IFI was defined as the recovery of yeast from blood culture or other normally sterile site. Fungemia was considered to be catheter-related if a catheter tip culture yielded the same yeast isolated in the bloodstream. Persistent fungemia was defined as the persistence of positive blood cultures for [2 days from the time of the first positive blood culture. Candida colonization was defined as repeated growth of yeasts from at least two different non-sterile sites. According to Pittet et al.'s definitions [14], the colonization index (CI) was defined as the ratio of the number of distinct non-blood body sites colonized by Candida to the total number of body sites cultured. Patients with a CI C0.5 were considered to have multifocal colonization.
Mold IFI was defined as filamentous fungi (e.g., Aspergillus, Fusarium, Mucormycetes) isolation from a normally sterile or non-sterile body site, in conjunction with suggestive clinical manifestations and instrumental test findings.
Corticosteroid treatment was defined as exposure to C10 mg/day prednisone equivalent for C30 days. Severity on admission was defined by the APACHE II score and organ dysfunction on IFI diagnosis was computed by the Sequential Organ Failure Assessment (SOFA) score [15]. Outcome was defined as survival or death within 30 or 60 days from the incident episode, in yeast and mold IFIs, respectively.

Laboratory procedures
All biological samples were cultured on two Sabouraud chloramphenicol dextrose agar plates (bioMèrieux, Marcy l'Etoile, France), incubated at 36 ± 1 and 28°C (for yeasts and molds isolation), and examined daily until 15 days.
Yeasts identification was performed with sugar assimilation profiles obtained using the ID32C kit (bioMérieux, Marcy l'Etoile, France). Filamentous fungi were identified at the levels of genera and species using macro-and micromorphology observations, according to standard methods [16]. GM in serum and bronchoalveolar lavage (BAL) was measured using a sandwich ELISA (Platelia Aspergillus Ag, Bio-Rad, Marnes La Coquette, France). An optical density ratio C0.5 in serum and C1 in BAL was considered to be positive. Samples that yielded positive results, in which interference was known to have occurred, were excluded.
Antifungal susceptibility tests were performed for yeast and mold isolates. The following drugs were supplied by the manufacturers as pure standard compounds: anidulafungin (AND), fluconazole (FLC), and voriconazole (VRC) (Pfizer Pharmaceuticals, Groton, CT, USA); caspofungin (CSP) and posaconazole (PSC) (Merck & Co., Inc., Whitehouse Station, NJ, USA); amphotericin B (AmB) (Sigma-Aldrich, Milan, Italy). The antifungal susceptibility was evaluated by broth microdilution assay performed according to the methodology recommended by the Clinical and Laboratory Standards Institute (CLSI) [17,18]. C. parapsilosis ATCC 22019, C. krusei ATCC 6258, Aspergillus fumigatus ATCC 204305, and A. flavus ATCC 204304 were used as quality controls and tested in each run of the experiments.
For Candida spp., the susceptibilities were interpreted taking into account the new species-specific clinical breakpoint suggested by the CLSI subcommittee [19].
AND and CSP minimum inhibitory concentration (MIC) values of B0.25 lg/mL were considered to be susceptible (S) for C. albicans, C. tropicalis; MIC results of B2 lg/mL were categorized as S for C. parapsilosis and C. guilliermondii; AND and CSP MIC end points B0.12 lg/mL were considered to be S for C. glabrata. C. albicans and C. tropicalis strains for which the echinocandin MIC was C1 lg/mL (C0.5 lg/mL for C. glabrata) are considered to be resistant (R). C. parapsilosis and C. guilliermondii strains for which the echinocandin MIC resulted C8 lg/mL were considered to be R.
FLC MIC end points B2, 4, and C8 lg/mL were categorized as S, susceptible dose-dependent (SDD), and R for C. albicans, C. tropicalis, and C. parapsilosis. FLC MIC values B32 and C64 lg/mL were considered to be SDD and R for C. glabrata.
For VRC, C. albicans, C. tropicalis, and C. parapsilosis were categorized as S in cases of MIC B0.125 lg/mL, intermediate (I) in cases of MIC 0.25-0.5 lg/mL, and R in cases of MIC C1 lg/mL. The epidemiological cut-off value (ECV) C1 lg/mL was used to detect resistance in C. glabrata [20]. The species-specific breakpoint for VRC was used for PSC.
Regarding AmB, in accordance to the literature data [21], a breakpoint B1.0 lg/mL was selected to define the isolates as S.
Clinical breakpoints have not yet been established for any antifungal agent against Aspergillus spp.; however, we applied ECVs suggested by Espinel-Ingroff et al. [22][23][24] to examine the susceptibility of this mold to the antifungal agents.

Statistical analysis
Continuous normally distributed data are expressed as mean and standard deviation (SD) and compared using unpaired Student's t-test. Non-normally distributed data are expressed as median and interquartile ranges (IQR) and compared using the Mann-Whitney U-test. Categorical data are expressed as number and percentage and compared using v 2 or the Fisher's exact tests. Logistic regression analysis was then performed, with Candida species as the dependent variable. Variables with a p-value \0.10 in the univariate analysis and those judged to be clinically relevant were included in the model. The power of the model was tested by the Hosmer-Lemeshow goodness-of-fit test. The effect of potential confounding factors was determined by introducing each factor independently in the final model and considering the variation in the model fit [25]. To evaluate the role of each variable as an independent risk factor, all variables associated with Candida species at the \0.05 level of risk in the logistic analysis were introduced in a backward stepwise logistic regression model with an a to remove of 0.05. In all comparisons, a p \ 0.05 was considered to be statistically significant. Data analysis was performed by using the Statistical Package for the Social Sciences (SPSS) software.

Results
Among 5,561 patients consecutively admitted to the participant ICUs during the 18-month survey, 105 patients with IFI were identified. The overall IFI incidence was 18.9 cases per 1,000 admissions. Yeasts and molds were responsible for 87. 6  As compared with patients with CnA BSI (Table 2), those with C. albicans BSI presented a higher incidence of comorbidities, specifically of diabetes mellitus, and were less frequently exposed to antifungal prophylaxis and to parenteral nutrition. Moreover, C. albicans BSI occurred earlier, but antifungal treatment was started later, with respect to CnA candidemic patients. In the logistic model, diabetes mellitus and abdominal surgery were significantly associated with C. albicans BSI (Hosmer-Lemeshow goodness-of-fit test, v 2 11.35, p = 0.91). In the stepwise analysis, diabetes mellitus was the only independent predictor of C. albicans BSI acquisition.
When compared to patients with Candida non-parapsilosis candidemia (Table 2), the cases of C. parapsilosis BSI presented less comorbidities, were more frequently admitted for trauma, and were exposed less frequently to antifungal prophylaxis and more frequently to parenteral nutrition. Moreover, these subjects manifested candidemia later during their ICU stay. In the binomial analysis, trauma admission diagnosis, length of ICU stay, and parenteral nutrition were promoting factors of C. parapsilosis candidemia, whereas antifungal prophylaxis manifested a protective role on its occurrence (Hosmer-Lemeshow goodness-of-fit test, v 2 10.52, p = 0.127). When these factors were fitted in the stepwise regression model, trauma, ICU LOS, and parenteral nutrition reached statistical significance, being the best independent promoting factors of C. parapsilosis BSI.  In all these patients, the chest X-ray showed suggestive findings (i.e., lobar consolidation, ill-defined nodules). Scedosporium apiospermum was responsible for one brain abscess. Computed tomography (CT) scan and cerebral magnetic resonance imaging (MRI) of this patient revealed an expansive left temporoparietal process with vasogenic edema.

Antifungal susceptibility testing
All 98 available isolates (92 yeasts, five A. fumigatus, and one S. apiospermum) were tested for in vitro antifungal susceptibility (Table 3).
Among yeasts, AmB and CSP had a susceptible rate of 94.6 and 95.5 %, respectively. All C. parapsilosis and C. tropicalis were AND susceptible, while a resistance was recognized in 2/37 C. albicans and 3/9 C. glabrata isolates.

Treatment
No patient with mold IFIs was on antifungal prophylaxis at the time of diagnosis, while 18 (19.6 %) candidemic patients were receiving FLC 200 mg/day, for a mean of 20.6 (±15.4) days. Candida spp. isolates from patients on antifungal prophylaxis were less susceptible to FLC, PSC, and VRC (Table 4). Table 3 In vitro antifungal susceptibilities (lg/mL) of 98 isolates collected from the 18-month survey, using Clinical and Laboratory Standards Institute (CLSI) methods Empirical therapy was started in 17.4 % of candidemic patients within 3 (±2) days from symptoms onset. Drugs empirically prescribed were FLC in 62.5 % of cases, followed by AmB (25 %) and CSP (12.5 %). In all cases, empirical therapy was confirmed as the target therapy. Overall, the target antifungal therapy was administered in 98 (91.8 %) patients; the remaining seven patients (five Candida spp., one A. fumigatus, and one S. apiospermum infections) were not treated because of pre-IFI diagnosis death.

Patient outcome
Overall, the crude mortality rate was 42.8 %. More specifically, yeast cause mortality was 40.2 %. Species-specific mortality ranged from 26.5 % for infection caused by C. parapsilosis to 77.8 % by C. tropicalis. CVC removal was associated with a better outcome [mortality rate, 63.6 vs. 27.1; odds ratio (OR) 0.21; 95 % confidence interval (CI) 0.08-0.58 p \ 0.05]; the exclusion of the patients who died within 48 h did not significantly modify this finding.
Autopsy was performed in one patient and showed disseminated candidiasis (kidney and liver).
Among molds, eight patients died (61.5 %); of these, two died (including the scedosporiosis case) as a direct consequence of infection (attributable mortality rate 15.4 %).

Discussion
Our multicenter survey shows that IFIs represent an important infectious complication in critically ill patients. A high occurrence of cases due to CnA (59.8 %) and a high mold mortality rate (61.5 %) were observed.
The most frequent IFI was Candida BSI, with an incidence of 16.5 cases per 1,000 admissions. This finding is higher than the data reported for Northern Europe (6.7 cases to 7.4 per 1,000 admissions) [1,26] and lower than those reported in other European countries (35.7-54 cases per 1,000 admissions) [2,27]. In an Italian study, Tortorano et al. [11] found an incidence of 10.08 per 1,000 admissions. The differences among various geographical areas may be due to several factors, including diversity in patient age, variations in clinical and microbiological practices, and differences in drugs usage [28]. CVCs are widely regarded as the most common risk factor for candidemia [3], and their removal is associated with shorter duration of disease and better outcome [29]. In the present study, 96.7 % of patients with candidemia had a catheter in place at the time of diagnosis. We found that all microbiologically evaluated catheters were probably the source of Candida BSI, and their removal was associated with higher survival rate (72.9 %).
Although C. albicans is still regarded as the most common species [3], recent epidemiologic studies have demonstrated an increasing incidence of CnA candidemia, with C. glabrata and C. parapsilosis ranked as second in the USA and Northern Europe [4,5], and in Latin America [30] and Southern Europe [8], respectively. The reason for this change in the pattern of Candida species distribution has not yet been completely understood, but some predisposing factors have been identified, such as indwelling catheters and parenteral nutrition for C. parapsilosis [7,31], cancer and neutropenia for C. tropicalis [32], and previous exposure to azoles for C. krusei and C. glabrata [33,34]. We found that CnA were the most frequent etiologic agents of candidemia, with C. parapsilosis ranked first (61.8 %). In agreement with other studies [2,31], we observed an association between ICU LOS, parenteral nutrition, and C. parapsilosis BSI. In reality, C. parapsilosis is notorious for its capacity to grow in hyperalimentation solutions with high concentrations of glucose, to form biofilms on catheters and other implanted devices, for nosocomial spread by healthcare workers' hands, and for its persistence in the hospital environment [31]. These observations suggest the need for more accurate nosocomial surveillance measures, such as hand hygiene, to prevent C. parapsilosis BSI and improve the health conditions of patients at risk.
Antifungal resistance was poorly recognized in our study and restricted to a few isolates. According to other authors [35,36], CSP and AND resistance was low (4.4 % to CSP and 5.6 % to AND). FLC still tended to be quite active against isolates of C. albicans and C. parapsilosis, while its resistance was mainly associated with C. glabrata and C. tropicalis. In our study, 8 out of 10 episodes of candidemia caused by FLC-resistant strains (five C. glabrata and three C. tropicalis) occurred in individuals with previous FLC administration. In addition, higher VRC and PSC MICs tended to be associated with FLC prophylaxis, confirming the potential problem of cross-resistance between azoles [37]. Finally, in our study, the mold infections were less common than yeast infections (ratio 1:7), but accounted for a higher mortality rate (61.5 vs. 40.2 %). Actually, IPA has recently gained importance in the ICU setting, ranging from 0.3 to 6.9 % [9,38], with an overall mortality rate of 80 % and an attributable mortality of 20 % [39]. It commonly occurs in patients with acute exacerbation of chronic obstructive pulmonary disease, diabetes, and in recipients treated with intravenous corticosteroids [9,11,40]. IPA diagnosis is difficult because of non-specific signs and symptoms and due to additional diagnostic examinations often delayed by a poor clinical suspicion [40,41]. In our study, an IPA incidence of 2.1/1,000 admissions and steroids employed in 76.9 % of these patients were observed. In this setting, the outcome was poor (60 %), in accordance with other reports describing dramatic fatality rates [9,41].
In conclusion, candidemia is the most frequent IFI in ICU patients (87.6 %), while molds IFI remains a sporadic event (12.4 %). CnA is responsible for over half of the candidemia episodes, with C. parapsilosis being the most common (61.8 %). A prolonged ICU LOS, use of invasive procedures (i.e., CVC, parenteral nutrition), and/or inadequate control of fungal infections could explain the high prevalence of C. parapsilosis. Mortality remains high (crude mortality 42.8 %), mainly linked to mold infections. Periodic surveillance is necessary in order to estimate the incidence of IFI as well the antifungal drug resistance. Further studies in the ICU will be needed so as to understand more clearly the interaction between fungi and host conditions, to allow earlier and more accurate diagnosis, and minimize the inappropriate use of drugs.