Introduction

Invasive Candida infections are the most common invasive fungal infections, accounting for 70-90% of all invasive mycoses [1]. Among the causes of nosocomial bloodstream infections, Candida ranks number four in the United States [2]. Incidence of candidemia varies between 0.5 and 1.4 per 10,000 patient-days in hospital and between 2 and 6.9 per 1,000 admissions in the intensive care unit (ICU) [3–7]. Invasive candidiasis is associated with high mortality, especially in ICUs [8–10]. Prompt initiation of appropriate antifungal therapy is essential for the control of invasive Candida infections and has been shown to reduce mortality [11–13].

Unfortunately, early diagnosis of invasive candidiasis remains a challenge, and criteria for starting empirical antifungal therapy in ICU patients are poorly defined. Recent IDSA guidelines suggest that "empirical antifungal therapy should be considered in critically ill patients with risk factors for invasive candidiasis and no other known cause of fever" [14]. Risk factors for invasive candidiasis are well identified [15]. However, these are so numerous that most ICU patients could be considered as exhibiting risk factors for invasive candidiasis. Obviously, widespread use of antifungal agents would be associated with substantially increased overall health care costs and emergence of resistance [16, 17].

To both ensure appropriate and timely antifungal therapy and to avoid unnecessary use of antifungal agents, some authors have developed clinical prediction rules to identify ICU patients at high risk of candidiasis and for whom initiation of empirical antifungal therapy could be justified [18–20]. However, there are many concerns about these rules: high specificity but low sensitivity, no prospective validation, and complicated use.

In 2006, a Spanish group, using the database of the Estudio de Prevalencia de CANdidiasis project, identified four predictors of proven invasive Candida infection [21]. Based on these predictors, a score named "Candida score" was built. In 2009, the same group demonstrated a significant linear association between increasing values of the "Candida score" and the rate of invasive Candida infections [22]. Such a score could be useful to stratify the risk of proven Candida infection and differentiate patients who would benefit from early antifungal treatment from those for whom invasive candidiasis is highly improbable.

The goal of the present study was to evaluate, in a prospective and multicenter cohort of ICU patients developing hospital-acquired severe sepsis or septic shock, the performance of the "Candida score."

Materials and methods

Study population

This prospective observational cohort study was performed from January 2010 to March 2011 in five ICUs (Lens, Lille, Maubeuge, Roubaix, and Tourcoing) in Nord-Pas de Calais, an area from the north of France. The study protocol was submitted to the Institutional Review Board for Lille University Hospital, which gave an approval with waiver of informed consent, in agreement with French regulations concerning observational studies that do not modify existing diagnosis or therapeutic strategies.

All patients exhibiting, on ICU admission or during their ICU stay, hospital-acquired severe sepsis or septic shock could be included in this study. Exclusion criteria were community-acquired infections, age < 18 years, neutropenia defined as a total leukocyte count < 500/mm3, pregnant women and nursing mothers, and patients who were treated with antifungal drugs when severe sepsis or septic shock occurred.

Study design

The primary objective of this study was to determine the incidence of invasive candidiasis among included patients and to study the relationship between the presence of invasive candidiasis and the "Candida score" value at the onset of severe sepsis or septic shock. The secondary objective was to analyze the antifungal therapies instituted when severe sepsis or shock occurred and to determine their relationship with the "Candida score" value.

Data collection, evaluation, and definition

On ICU admission, demographic characteristics, underlying diseases, reason for ICU admission, and severity of illness were recorded by each investigator on a standardized report form. Underlying diseases included insulin-dependent diabetes mellitus, neurologic conditions, chronic obstructive pulmonary disease, chronic liver disease, chronic renal failure, congestive severe heart failure, and immunosuppression (prednisolone >0.5 mg/kg/d since more than 1 month, human immunodeficiency virus infection, cancer chemotherapy within the past 3 months, organ transplantation with ongoing immunosuppressant, bone marrow allograft or hematopoietic stem-cell transplantation, immunosuppressant, or anti-tumor necrosis factor).

According to diagnoses at the time of ICU admission, patients were classified as surgical (admission for the postoperative care of an elective or urgent surgical procedure), trauma (admission because of trauma-related acute lesions), or medical (all other patients). Severity of illness on admission in ICU was assessed by using the Simplified Acute Physiology Score (SAPS) II and the Sepsis-related Organ Failure Assessment (SOFA) score [23, 24].

During the ICU stay, screening for Candida colonization was performed twice weekly by routine samples from tracheal aspirates and urine. Other samples from peripheral blood, vascular catheters, wound or drainage exudates, or other infected foci could be obtained at the discretion of the attending physician. Samples were processed by the different clinical microbiology laboratories of the participating hospitals.

When hospital-acquired severe sepsis or septic shock occurred, components of the "Candida score" and therapeutic data were collected. Severe sepsis and septic shock were defined according to international sepsis definition [25]. They were considered as hospital-acquired when infection was neither present nor incubating at the time of admission to the acute care setting [26]. Components of "Candida score" were severe sepsis, total parenteral nutrition, surgery, and multifocal Candida colonization. These components were defined according to criteria proposed by Leon et al. [21]. The "Candida score" was calculated by adding points provided by each component as proposed by Leon et al. in their second study [22]. In our cohort, the minimum value of the score was 2 points for all patients, because severe sepsis or septic shock were inclusion criteria. The total score was obtained by adding 1 point for each remaining variable, total parenteral nutrition, surgery, and multifocal Candida colonization. Candida colonization was considered multifocal when Candida species were isolated, at the same moment, from two or more noncontiguous foci, even if Candida species were different. In some patients, data about Candida colonization were unknown when severe sepsis or septic shock occurred (i.e., patients admitted in ICU from another ward or another hospital). Samples from tracheal aspirates or urine were obtained at admission and the final Candida score was only determined when cultures results were available. Finally, the decision to treat a patient with antifungal drugs during the course of this observational study was at the discretion of each attending physician. In all cases, the date of starting antifungal treatment and antifungal agents administered were recorded.

In all patients, usual risk factors for invasive Candida infection, not included in the "Candida score," such as exposition to invasive devices (endotracheal tube for mechanical ventilation, central venous catheter, implantable drug delivery system, urinary catheter), antibiotic therapy for more than 5 days within the past 2 weeks, immunosuppression, renal failure, and insulin-dependent diabetes mellitus were recorded.

Invasive candidiasis was defined according to usual criteria, such as those proposed by Leon et al. [21, 22]. Briefly, candidemia was defined by at least one positive blood culture, and peritonitis was diagnosed on the basis of macroscopic findings and direct examination or positive culture for Candida of the peritoneal fluid collected during surgical procedure. Isolation of Candida species from normally sterile body fluids, such as pleural or pericardial fluid, also was considered as definite criteria of invasive candidiasis. Patient mortality at the time of ICU discharge was determined.

Statistics

Variables were expressed as median values and ranges for numerical variables and as frequencies and percentages for categorical variables. Categorical variables were compared using the Chi-square and Fisher's exact tests. Numerical variables were compared using parametric test (Student's test) or nonparametric tests (Wilcoxson, Kruskal-Wallis) according to the sample size of the groups, the number of groups, and the normality of parameters. Statistical significance was accepted at the 5% level.

Results

Ninety-four patients were included. Main patients' characteristics on ICU admission are summarized in Table 1. At inclusion, 73 patients exhibited septic shock and 21 had severe sepsis. The median time between ICU admission and occurrence of shock or severe sepsis was 7 (range, −1 to 37) days. Mean SOFA score was 9.9 ± 2.12.

Table 1 Main patients' characteristics on ICU admission
Full size table

When severe sepsis or shock occurred, 27 patients had a total parenteral nutrition, 31 had undergone a surgical procedure immediately before or during ICU stay, and 17 had a multifocal Candida colonization. Candida species were found in urine (n = 17) and tracheal aspirates (n = 17) cultures. Thirty-seven Candida isolates were identified: C. albicans (n = 30), C. tropicalis (n = 3), C. parapsilosis (n = 2), and other species (n = 2). Calculated and distribution of "Candida scores" was as follows: 44 patients with no risk factor had a score = 2; 29 patients with a lone risk factor (total parenteral nutrition n = 8, surgery n = 10, and multifocal Candida colonization n = 11) had a score = 3; 17 patients with two risk factors (total parenteral nutrition n = 15, surgery n = 17, and multifocal Candida colonization n = 2) had a score = 4. Finally, four patients had all risk factors and a score = 5.

Numerous other risk factors for invasive Candida infection were present and are reported in Table 2, broken by "Candida score" value. Among the 94 patients, severe sepsis or septic shock were due to or associated with invasive candidiasis in 5 (5.3%) patients. One patient had C. parapsilosis candidemia, three patients had C. albicans peritonitis, and one patient had C. tropicalis pleural infection. The patient with C. parapsilosis candidemia had C. albicans-positive tracheal aspirates cultures and C. parapsilosis- positive urine cultures. Among the three patients with peritonitis, only one had multifocal Candida colonization with C. albicans- positive tracheal aspirates and urine cultures. The patient with C. tropicalis pleural infection had C. albicans-positive tracheal aspirates and urine cultures.

Table 2 Risk factors for invasive candidiasis, according to the value of "Candida score"
Full size table

The rates of invasive candidiasis according to the "Candida score" was 0% in the 44 patients with score = 2 and the 29 patients with score = 3, 17.6% (n = 3) in the 17 patients with score = 4, and 50% (n = 2) in the 4 patients with score = 5. The association between increasing values of the "Candida score" and the rate of invasive candidiasis is statistically significant (p < 0.0001). In patients with a "Candida score" ≤3, no invasive candidiasis was observed. The positive and negative predictive values of this cutoff value of the "Candida score" were 23.8% and 100%, respectively. Among the 68 patients with medical diagnosis on ICU admission, 4 had a "Candida score" >3 and 1 of 4 exhibited an invasive candidiasis. Among the 25 patients with a surgical diagnosis on ICU admission, 17 had a "Candida score" >3 and 4 of 17 had an invasive candidiasis. Although the number of patients with a "Candida score" >3 was higher in surgical than medical patients (p < 0.0001), the rate of invasive candidiasis was similar for surgical and medical patients with a "Candida score" >3 (4/17 vs. 1/4; p = 1). When severe sepsis or septic shock occurred, 19 (20%) patients received empiric antifungal treatment with fluconazole in 12 cases and caspofungin in 7 cases. Empiric antifungal treatment was initiated in 1 of 44 (2.3%) patients with a "Candida score" = 2, 8 of 29 (27.6%) patients with a "Candida score" = 3, 7 of 17 (41.2%) patients with a "Candida score" = 4, and 3 of 4 (75%) patients with a "Candida score" = 5. All five patients with invasive candidiasis received, when severe sepsis or septic shock occurred, empirical antifungal therapy.

Forty-six (49%) patients died during their ICU stay. Among the 68 patients with a medical diagnosis on ICU admission, 32 (47%) died, whereas among the 25 patients with a surgical diagnosis on ICU admission, 14 (56%) died (p = 0.45). Mortality rates according to the values of the "Candida score" and the empiric initiation of antifungal treatment are reported in Table 3. No statistically significant difference was observed. Among the 73 patients with a "Candida score" ≤3, 33 (45%) died, whereas among the 21 patients with a score >3, 13 (62%) died (p = 0.17). An empiric antifungal treatment was instituted in 19 patients, and 10 (53%) of them died. Thirty-six (48%) of the 75 patients who did not receive empiric antifungal treatment died. Such a difference was not statistically significant (p = 0.72). Finally, among the 21 patients with a "Candida score" >3, mortality was not different for patients who received or did not receive empiric antifungal treatment (6/10 vs. 7/11; p = 1).

Table 3 Severity scores and mortality, according to the value of "Candida score"
Full size table

Discussion

This prospective cohort study confirms that the "Candida score" is an interesting tool to differentiate, among ICU patients with hospital-acquired severe sepsis or septic shock, those who would benefit from early antifungal treatment (score > 3) from those for whom invasive candidiasis is highly improbable (score ≤ 3).

For patients who develop, during ICU stay, severe sepsis or septic shock, it has been demonstrated a few decades ago that prompt and adequate antimicrobial treatment is associated with prognosis improvement. Moreover, it was demonstrated that bloodstream infections due to Candida species were independently associated with administration of inadequate antimicrobial treatment [27]. These results led physicians to question the interest of adding an antifungal agent to the empirical antibiotic regimen administered to patients with severe sepsis or septic shock. However, criteria for starting empirical antifungal therapy in ICU patients are poorly defined and recent IDSA guidelines suggesting that "empirical antifungal therapy should be considered in critically ill patients with risk factors for invasive candidiasis and no other known cause of fever" could lead to an overuse of antifungal agents. The "Candida score" could overcome the dilemma between overuse, associated with increased health costs and development of resistance, and underuse, associated with prognosis impairment of critically ill patients.

In 2006, Leon et al. reported a prospective, cohort, observational, multicenter study of 1,669 adult ICU patients admitted to 73 medical-surgical Spanish ICUs between May 1998 and January 1999 [21]. They identified four independent factors associated with a greater risk for proven invasive Candida infection. A bedside scoring system called "Candida score" was built as follows: Candida score = 0.908 × (total parenteral nutrition) + 0.997 × (surgery) + 1.112 × (multifocal Candida species colonization) + 2.038 × (severe sepsis). All variables were coded 1 if present and 0 if absent. According to receiver operating characteristics (ROC) curve and area under the ROC curve, a score > 2.5 accurately identified patients with the greater risk of proven invasive Candida infection (sensitivity 81%, specificity 74%). In 2009, Leon et al. prospectively studied 1,107 adult ICU patients admitted between April 2006 and June 2007 in 36 medical-surgical ICU in Spain, France, and Argentina [22]. They used a rounded "Candida score" calculated as follows (all variables were coded 0 if absent and 1 if present): 1 × (total parenteral nutrition) + 1 × (surgery) + 1 × (multifocal Candida colonization) + 2 × (severe sepsis). They demonstrated a linear and significant association between increasing values of the "Candida score" and the rate of invasive candidiasis. The incidence rate was 2.3, 8.5, 16.8, and 23.6 when the score was <3, 3, 4, and 5, respectively. They concluded that invasive candidiasis is highly improbable in patients with "Candida score" <3.

Our results are in accordance with Leon's data. The incidence rates of invasive candidiasis were 0%, 0%, 17.6%, and 50% in patients with scores equal to 2, 3, 4, and 5, respectively. A linear and significant association between increasing values of the "Candida score" and the rate of invasive candidiasis was observed, and no invasive candidiasis occurred in patients with scores <3. Moreover, no invasive candidiasis was observed in patients with a "Candida score" = 3. Interestingly, we also can underline that the performances of this score are similar in medical and surgical patients. Surgical patients exhibit more often a "Candida score" >3 than medical patients, but the rate of invasive candidiasis is similar in surgical and medical patients with a "Candida score" >3.

Therapeutic decisions of intensivists in our group could be discussed. Decision to initiate antifungal drugs during this observational study was at the discretion of each attending physician. Nevertheless, we observed a relationship between initiation of antifungal agents and value of the "Candida score." Frequency of antifungal empirical therapy were 2.3%, 27.6%, 41.2%, and 75% in patients with scores equal to 2, 3, 4, and 5, respectively. In our opinion, this is rather satisfying because most patients with highly improbable invasive candidiasis did not receive antifungal drugs, whereas early empirical antifungal treatment was initiated in most patients who would benefit from it. However, the fact that only a quarter of patients with score = 3 received an empiric antifungal treatment merits some comments. In our series, all included patients exhibited severe sepsis or septic shock and, therefore, had a "Candida score" always at least equal to 2. Initiation of empirical antifungal treatment to all patients with another risk factor (surgery or total parenteral nutrition or multifocal Candida colonization) probably seemed to be excessive to most intensivists of our group. The absence of invasive candidiasis, among the 29 patients with a "Candida score" = 3, retrospectively confirm the appropriateness of their decision.

This study has numerous limitations. First, the number of included patients is low compared with the number of patients studied by Leon and his group [21, 22]. The inclusion of patients with severe sepsis or septic shock rather than all ICU admitted patients, and the low number of ICU participating to this cohort study explain this fact. Second, only five invasive candidiasis and, among them, only one candidemia were observed in this cohort. The low number of invasive candidiasis can be explained, first, with some exclusion criteria such as neutropenia, and, second, with the setting of the ICU; four of them were in a secondary care hospital and only one in a university hospital. Third, in some patients, the "Candida score" could not be calculated when sepsis or shock occurred, because some mycological data were lacking. However, this only reflects the shortcomings of a real-life study and represents one limit to the practical use of the score.

Conclusions

Our study confirms the clinical relevance of the "Candida score" in patients with hospital-acquired severe sepsis or septic shock. This score can differentiate patients who would benefit from early antifungal treatment from those for whom invasive candidiasis is highly improbable.