Introduction

Candida spp. colonization occurs in up to 80 % of critically ill patients after 1 week in intensive care [13]. This very high proportion contrasts strongly with the low rate of invasive candidiasis in less than 10 % of them [4, 5]. Despite recent advances in microbiological techniques, early diagnosis of invasive candidiasis remains problematic and microbiological documentation often occurs late in the course of infection [68]. This explains its high crude and attributable mortality, i.e., in the range reported for septic shock [911].

Early empirical treatment of severe candidiasis has improved survival, but is responsible for the overuse of antifungals [1214]. This overuse not only contributes to a huge financial burden, but has also promoted a shift to Candida species with reduced susceptibility to antifungal agents [15, 16]. Unfortunately, recent guidelines resulting from expert consensus failed to provide high-level recommendations about empirical antifungal treatment or to clarify the nature of such treatment strategies [8, 17, 18]. Despite limited levels of evidence, empirical treatment currently relies on the positive predictive value of risk assessment strategies, such as the colonization index (CI), Candida score, and predictive rules based on combinations of risk factors [1921].

Improved knowledge of the pathophysiological specificities of invasive candidiasis should promote better use of clinical tools in the evaluation of patients who could truly benefit from early antifungal therapy [22]. In contrast to most bacterial infections, invasive candidiasis is characterized by a 7–10-day delay between exposure to risk factors and infection development [2325]. This time window provides a unique opportunity for the implementation of a structured approach to the rigorous selection of patients likely to benefit from early empirical antifungal treatment [19, 26].

The CI is the most widely studied clinical tool for the early risk assessment of invasive candidiasis among at-risk patients. The primary objective of this review is to highlight its role in this setting. The use of colonization dynamics, as assessed by the CI, for the early detection of patients likely to benefit from early antifungal treatment is also explored in comparison with other tools currently available in clinical practice. Finally, we also propose a research agenda for the next decade.

Pathophysiology of invasive candidiasis

Nosocomial exogenous transmission of Candida has been well described [27]. Of note, both endogenous and exogenous colonization can co-exist in the clinical setting [1]. However, carefully designed studies using genotyping of Candida strains confirmed that endogenous colonization is responsible for the large majority of severe candidiasis [28, 29].

The epidemiology of invasive candidiasis has some important differences in immunocompromised and critically ill patients (Fig. 1).

Fig. 1
figure 1

Pathophysiology of invasive candidiasis in different patient populations. Detailed mechanisms are summarized in the corresponding sections of the text

Neutropenic and bone marrow transplant recipients

Neutrophils are essential in the defense against invasive candidiasis [30]. Their prolonged absence in neutropenic patients or those with functional impairment after bone marrow transplantation, combined with the selective pressure of frequent and repetitive exposure to antibacterial agents, plays a major role in the development of invasive candidiasis in these patients. A high density of Candida spp. on mucosal surfaces injured by chemotherapy develops progressively and is responsible for fungal translocation with a high risk of candidemia. The use of systematic antifungal prophylaxis in the past three decades explains the epidemiology of Candida infection, typically characterized by breakthrough invasive candidiasis [31, 32]. Infection is almost always caused by a Candida strain resistant to the antifungal agent used [15, 33].

Solid organ transplant recipients

Anti-rejection therapy increases the risk for Aspergillus and other filamentous fungal infections in solid organ transplant recipients, except among those receiving small bowel and liver transplants, in whom additional specific surgical factors further modify the epidemiology of disease [34, 35]. Prolonged and intense impairment of cellular T cell-mediated immune response may contribute to the high proportion of cutaneous and mucosal candidiasis [36]. This factor has justified the use of systematic antifungal prophylaxis in these patients in whom, as among neutropenic patients and bone marrow transplant recipients, breakthrough Candida spp. infection develops with strains resistant to the agent used [18].

Nonsurgical critically ill patients

In nonsurgical critically ill patients, the continuous and prolonged support of failing organs and the selective pressure of broad-spectrum antibiotics constitute key risk factors for invasive candidiasis. Support of organ failure requires the use of numerous devices, such as intravascular catheters, endotracheal tubes, naso- and oro-gastric tubes, and Foley catheters, which are frequently colonized by Candida spp. as a result of the high affinity of their biofilm [1]. These specificities may explain progressive colonization in a high proportion of patients after prolonged stay in the intensive care unit (ICU) [6, 37, 38]. They may also explain a higher proportion of catheter-related infections in the absence of severe immune impairment [39]. In this context, the role of exposure to steroids largely used to improve the speed of recovery from circulatory failure remains to be determined.

Limited data are available on the usefulness of prophylaxis and recent guidelines did not include recommendation for antifungal prophylaxis in nonsurgical critically ill patients [17, 18, 40].

Critically ill patients recovering from abdominal surgery

Additional factors should be considered in patients after abdominal surgery [41]. Any perforation or opening of the digestive tract results in contamination of the peritoneum by bowel flora. In most cases, surgical cleaning of the abdominal cavity, eventually combined with antibiotics, is sufficient to allow full recovery [42]. Except for patients presenting with nosocomial peritonitis and some of those presenting with septic shock and multiple organ failure, the identification of Candida spp. has no clinical significance under these conditions [22]. Recurrent peritonitis following anastomotic leakage or persistent inflammation may be required for the progression of Candida spp. colonization to invasive candidiasis [43]. However, the development of invasive candidiasis requires concomitant exposure to additional factors, such as an increasing amount of Candida spp. in a non-functioning bowel, prolonged antibiotic treatment, and/or requirement for organ support. In contrast to nonsurgical patients, these particularities explain why a high proportion of invasive candidiasis cases do not manifest with candidemia and develop only late in the course of the disease [44].

Non-candidemic invasive candidiasis includes intra-abdominal abscess and peritonitis defined by one of the following culture results from specimens obtained at surgery: (1) monomicrobial growth of Candida spp.; (2) any amount of Candida spp. growth within a mixed-flora abscess; or (3) moderate or heavy growth of Candida spp. in mixed-flora peritonitis treated with appropriate antibacterial therapy according to susceptibility testing [4346].

Two small prospective studies, including one placebo-controlled, suggested that antifungal prophylaxis in patients presenting with anastomotic leakage after abdominal surgery may prevent the development of invasive candidiasis [18, 45, 46].

In patients with severe acute necrotizing pancreatitis, invasive candidiasis develops with similar pathophysiological characteristics. Progressive colonization of the bowel within the first 2 weeks of the disease results in translocation into necrotic tissues, and fungal infections have been documented in up to 10 % of patients not exposed to antibiotics.

Critically ill patients with Candida isolated from the respiratory tract

Candida spp. colonization of the airway is frequently reported in mechanically ventilated critically ill patients, and its clinical significance is difficult to evaluate. Candida has a low affinity for alveolar pneumocytes and histologically documented pneumonia has been rarely reported. Hematogenous dissemination in the context of candidemia [1] may be responsible for multiple pulmonary abscesses and should be viewed as a distinct entity. Hence the existence of true candidal pneumonia is doubtful and recovery of Candida spp. from the respiratory tract should generally be considered as colonization and does not justify antifungal therapy [1719].

Candida colonization

The distinction between patients with Candida spp. colonization who did not require treatment and those likely to develop invasive infection was identified as a difficult clinical challenge in the field of infectious diseases more than 40 years ago [47]. After the description in the early 1970s of Candida infection in surgical patients, manifesting as “disseminated candidiasis” and characterized by persistent fungemia and multiple disseminated microabscesses on autopsy, several authors reported that it might be preceded by fungal colonization. Klein and Watanakunakorn [48] reported the presence of visceral microabscesses in up to 30 % of patients with Candida spp. colonization and recommended treatment of all such patients with critical illness or persistent candidemia.

Solomkin et al. confirmed this concept in a detailed review of 63 surgical patients with postoperative candidemia. Among 51 patients who developed candidemia as a late complication of intra-abdominal infection, adequate antifungal treatment (>3 mg/kg amphotericin B) resulted in survival of 10/15 patients compared with 6/36 untreated patients. Autopsies revealed visceral Candida spp. microabscesses in 7/20 patients [42, 49]. Moreover, a digestive fungal source was identified in nearly all patients, and candidemia was considered to originate from intestinal Candida spp. growth in the presence of mucosal loss.

Calandra et al. provided further evidence of the clinical significance of Candida colonization of the peritoneum in surgical patients [39]. In a series of 49 patients with spontaneous perforation (n = 28) or surgical opening of the digestive tract (n = 21), 19 (39 %) patients developed invasive candidiasis, including 7 cases of intra-abdominal abscesses and 12 cases of peritonitis [43]. Invasive candidiasis developed significantly more frequently in patients who had undergone surgery for acute pancreatitis than in those with gastrointestinal perforations (90 vs. 32 %, p = 0.005) or other disorders, such as diseases of the biliary tract or colon (90 vs. 17 %, p = 0.003). Invasive candidiasis was associated significantly with an initially high or increasing amount of Candida spp., as assessed by semiquantitative cultures. Compared with uninfected patients, Candida spp. showed light, moderate, and heavy growth from the first positive specimen in 26 (87 %) vs. 9 (47 %), 4 (13 %) vs. 6 (32 %), and 0 vs. 4 (21 %) infected patients, respectively (p = 0.005). Moreover, the amount of Candida spp. cultured from subsequent specimens increased in 15 (79 %) infected patients and only 2 (7 %) uninfected patients (p < 0.001) [39].

Candida colonization index: the missing link

In the early 1990s, the origin of Candida spp. infection was highly debated [50]. Whether the infection was exogenous or endogenous in origin, and whether patients could become colonized with strains from the surrounding environment or with their “own” strains of Candida spp. was uncertain. We tested the hypothesis that yeast colonization would precede infection in critically ill patients [23]. We determined the genotypic characteristics of 322 Candida spp. collected prospectively from 29 critically ill patients who developed significant colonization under routine culture surveillance two to three times weekly over a 6-month period. These patients belonged to a cohort of 650 surgical ICU patients. Significant colonization was defined as the presence of Candida spp. in three or more samples taken from one or more sites on at least two consecutive screening days. Candida spp. strains isolated from an individual patient had an identical genetic pattern, even when isolated from different body sites, and this pattern remained the same over a prolonged period of up to 140 days. No horizontal transmission could be demonstrated during the study period. Invasive candidiasis, including 8 cases of candidemia, occurred in 11/29 (38 %) patients. All patients who developed infection had previous colonization by strains with identical genetic patterns. Our intensive surveillance failed to document cross-transmission of Candida spp. between colonized and non-colonized patients.

Colonization index

The development of the CI has been viewed as a major conceptual advance in the characterization of supporting the progression from colonization to infection in surgical patients.

The Candida CI is defined as the ratio of the number of distinct non-blood body sites colonized by Candida spp. to the total number of body sites cultured [24]. With the exception of blood cultures, samples collected from body sites other than those routinely screened are also considered in the CI (Table 1). Only strains of Candida spp. with the same genetic identity are considered in calculating the CI.

Table 1 Description of the colonization index

All 29 included patients were heavily colonized with Candida spp., but with no significant difference in the reason for ICU admission and comorbidities between infected and uninfected patients. The Acute Physiology and Chronic Health Evaluation II (APACHE II) score and duration of antibiotic exposure before colonization were higher among the 11 patients who ultimately developed invasive candidiasis. In these patients, colonization preceded infection by an average of 25 days (range 6–70). The 18 patients who did not develop infection were colonized for a mean of 29 days (range 5–140). A total of 153 [5.3 ± 1 (range 3–8) per patient] distinct body sites were tested, and this number did not differ significantly between the two groups of patients.

The CI was calculated daily, and the highest values obtained before invasive candidiasis development were compared with those recorded for patients who did not develop candidiasis. The average CI differed significantly between colonized and infected patients (0.47 vs. 0.70, p < 0.01). All patients who ultimately developed infection reached the threshold value of 0.5 before infection (Fig. 2). Importantly, infected patients reached the threshold CI value an average of 6 days (range 2–21) before candidiasis. All but one patient reached this threshold at least 3 days before the time of infection. The sensitivity, specificity, and positive and negative predictive values of the CI were determined to be 100, 69, 66, and 100 %, respectively.

Fig. 2
figure 2

Colonization index. This index is defined as the ratio of the number of distinct non-blood body sites with Candida spp. colonization to the total number of distinct body sites tested. It was recorded for each patient from the first day of colonization until discharge from the intensive care unit (uninfected patients) or severe candidiasis development (infected patients). Black circles represent patients who developed severe candidiasis, white circles represent patients who remained colonized. Reproduced with permission from Pittet et al. [24]

Corrected colonization index

In a post hoc analysis, we developed the corrected colonization index (CCI) by taking into account the amount of Candida spp. recovered by semiquantitative cultures [24]. The CCI is the product of the CI and the ratio of the number of distinct body sites showing heavy growth (+++ or at most 105 in urine or gastric juice) to the total of distinct body sites with Candida spp. growth (Table 1).

The mean CCI differed significantly between colonized (0.16) and infected (0.56) patients (p < 0.01). CCIs were less than 0.35 in all colonized patients and at least 0.4 in all infected patients (p ≤ 0.001). The sensitivity, specificity, and positive and negative predictive values were all 100 %. In a multiple logistic regression analysis, only the APACHE II score [odds ratio (OR), 1.03; 95 % confidence interval, 1.01–1.05 per point; p = 0.007] and CCI (OR, 4.01; 95 % confidence interval, 2.16–7.45; p < 0.001) independently predicted the development of invasive candidiasis.

Overview of the usefulness of the colonization index

Since the publication of the paper describing the CI in 1994 in Annals of Surgery, many centers have used the CI or a methodology derived from its original description to assess the dynamics of Candida colonization in different subgroups of critically ill patients at risk of invasive candidiasis. Unfortunately, these data have not been validated in large multicenter trials.

Several studies have indirectly suggested the validity and potential usefulness of the CI, but almost exclusively in surgical patients (Table 2). This index has been used to characterize colonization dynamics [37, 38, 5155], assess the significance of candiduria in critically ill patients [38, 5659], and evaluate the impact of antifungal prophylaxis in subgroups of patients at risk of candidiasis [40, 46, 6063]. At least four studies have reported the use of the CI to guide empirical antifungal treatment. Although major methodological flaws may preclude the validation of their findings, this application reduced the incidence of ICU-acquired invasive candidiasis in all four studies (Table 3) [57, 6466].

Table 2 Reported experience with the use of the colonization index (CI)
Table 3 Use of the colonization index (CI) to guide the initiation of empirical antifungal treatment

Dubau et al. [57] prospectively monitored CIs weekly in 89/669 consecutive patients staying in the ICU for more than 7 days after digestive surgery who had protein C levels greater than 100 mg/l or received antibiotic treatment for more than 7 days. All patients developing a CI above 0.5 were empirically treated with antifungals until it decreased below 0.5. The proportion of patients with CIs greater than 0.5 increased from 6 to 25, 40, 55, and 59 % after 1, 2, 3, 4, and more than 4 weeks, respectively. Only one of the 35 patients with CIs greater than 0.5 who were empirically treated with antifungal medications developed candidiasis; the degree of colonization decreased rapidly in the other 34 patients.

In a before/after trial, Piarroux et al. [64] prospectively used weekly CIs and CCIs to assess the intensity of Candida spp. colonization in 478 surgical patients staying in the ICU. Patients with CCIs greater than 0.4 received empirical antifungal treatment. Compared with an historical cohort of 455 control subjects, the incidence of invasive candidiasis was lower among these patients (7.0 vs. 3.8 %; p = 0.03). Moreover, this strategy completely prevented the development of ICU-acquired invasive candidiasis. The proportions of patients treated empirically were 87 % of those with CCIs greater than 0.4 (some of them died before the CCI could be calculated), 18 % of those with CCIs less than 0.4 and CIs greater than 0.5, and 2 % of those with CCIs less than 0.4. A total of 25 % of patients received empirical antifungal treatment. The authors estimated that 160 complete mycological screenings and 10 preemptive antifungal treatments were needed to prevent one proven ICU-acquired invasive candidiasis. Despite the absence of randomization, this study is the most valuable ever performed using the CI to guide empirical antifungal treatment. However, the intense laboratory work required to perform CI in all ICU patients and the high proportion of patient treated may explain why this practice is not currently used widely among non-immunocompromised critically ill patients.

In a prospective study of 37 critically ill patients, Eren et al. [65] found a correlation between the CI and the presence of C. albicans antibodies. Among 26 patients with C. albicans colonization, immunoglobulin M (IgM) and IgG were found in 5/7 (71 %) patients with CIs greater than 0.5, compared with only 3/12 (25 %) patients with CIs less than 0.5 in whom serum could be tested (total, n = 19) and 0/7 non-colonized patients in whom serum could be tested (total, n = 11) [65]. The authors stated, “in the follow-up of the patients, no candidemia developed and this was thought to be due to the preventive measures which were taken especially in ICU patients with CI > 0.5,” suggesting that empirical antifungal treatment was given to patients with CIs greater than 0.5.

Comparison of the colonization index with other predictive tools

The accuracy of the CI has been compared with that of other surrogate markers of invasive candidiasis, such as the Candida score, predictive rules, and, more recently, beta-glucan (Table 4) [44, 6772]. In a prospective study of 89 high-risk surgical ICU patients (61 recurrent gastrointestinal tract perforation; 25 severe acute pancreatitis) in whom 29 developed a non-candidemic invasive candidiasis, Tissot el al. showed that two consecutive beta-glucan serum levels above 80 pg/ml predicted early the development of infection with a sensitivity, specificity, positive predictive value, and a negative predictive value of 65, 78, 68 and 77 %, respectively [44]. Similarly, a previous study by Posteraro et al. [70] in a cohort of medical and surgical critically ill patients staying more than 5 days and developing a severe sepsis, in whom beta-glucan was higher than 80 pg/ml, showed a sensitivity, specificity, positive predictive value, and a negative predictive value of 93, 94, 72, and 99 %, respectively for the early detection of invasive candidiasis (16 episodes, including 13 candidemia). Overall, the accuracy of CI is lower than other methods for the early detection of patients at higher risk of infection. Nevertheless, the usefulness of beta-glucan to guide empirical antifungal treatment remains to be determined.

Table 4 Comparison of the colonization index (CI) with other markers for the prediction of invasive candidiasis

Three industry-sponsored studies failed to demonstrate the clinical usefulness of predictive rules combining various risk factors. A predictive rule was used in a multicenter, randomized, double-blind, placebo-controlled trial comparing caspofungin vs. placebo as antifungal prophylaxis or preemptive treatment in 222 critically ill patients [73]. The rule combined ICU stay for at least 3 days, mechanical ventilation, exposure to antibiotics, the presence of a central line, together with at least one additional risk factor among the following: parenteral nutrition, dialysis, surgery, pancreatitis, systemic steroids, or other immune suppressive agents. The incidence of invasive candidiasis was 16.7 % (14/84) vs. 9.8 % (10/102) in patients receiving placebo and caspofungin, respectively (p = 0.14). When baseline infections were included, it was 30.4 % (31/102) and 18.8 % (22/117), for patients receiving preemptive placebo and caspofungin, respectively (p = 0.04). Safety, length of ICU stay, antifungal use, and mortality did not differ. The authors concluded that caspofungin prophylaxis was safe with a non-significant trend to reduce invasive candidiasis, and that the preemptive therapy approach deserves further study. Two recent and currently unpublished studies attempt to use clinical predictive rules to guide empirical antifungal treatment. The first, entitled “Pilot Feasibility Study With Patients Who at High Risk For Developing Invasive Candidiasis in a Critical Care Setting (MK-0991-067 AM1” (ClinicalTrials.gov identifier: NCT01045798) and conducted by the Mycoses Study Group, was terminated because of a low recruitment rate after the inclusion of only 15 patients and generation of uninterpretable data. Preliminary results of the second study, entitled “A Study to Evaluate Pre-emptive Treatment for Invasive Candidiasis in High Risk Surgical Subjects (INTENSE)” (ClinicalTrials.gov identifier: NCT01122368), showed a high proportion of invasive candidiasis at study entry; independent experts considered, however, that the majority of salvage antifungal treatments were unjustified. Globally, the proportion of invasive candidiasis developing in the group receiving preemptive antifungal treatment was not different from that of the group receiving placebo, but the number of patients excluded from the analysis resulted in an underpowered analysis.

These outcomes illustrate a well-known paradigm of accuracy. The use of the most sensitive test potentially increases the number of patients who are uselessly treated empirically. In contrast, the use of the most specific test potentially increases the number of patients who do not receive empirical antifungal treatment and in whom late treatment could be associated with a worse outcome.

An ongoing prospective multicenter, double blind, randomized-controlled French trial (EMPIRICUS) is currently recruiting critically ill patients at risk of invasive candidiasis. Patients mechanically ventilated for more than 4 days with sepsis of unknown origin and with at least one extra-digestive fungal colonization site and multiple organ failure are eligible for randomization to receive empirical micafungin or a placebo. This study includes prospective determination of the CI among its secondary endpoints and may help to develop guidelines for treating non-immunocompromised patients with fungal colonization multiple organ failure and sepsis of unknown origin [74].

The CI and other predictive tools (the Candida score and predictive rules) have been specifically developed by using their positive predictive value for the early identification of high-risk patients who will develop invasive candidiasis. Nevertheless, it is relevant to note that the negative predictive values of the CI and all other surrogate markers of invasive candidiasis are much higher than their positive predictive values [21]. Among them, only the negative predictive value of the Candida score has been validated in a multicenter prospective clinical trial [75].

Advantages and pitfalls of the use of colonization index

To summarize the data reviewed on the use of the CI, its main advantages are that it has been successfully used to characterize colonization dynamics, assess the significance of candiduria, and to evaluate the impact of antifungal prophylaxis. A low CI has a high negative predictive value for the further development of invasive candidiasis. Among the pitfalls, it should be emphasized that it is work-intensive with a limited bedside practicability, that only limited data are available for nonsurgical patients, and its cost-effectiveness and usefulness for the management of critically ill patients remain to be proven in large prospective clinical trials.

Further perspectives

These observations may pave the way for a new paradigm in the research agenda of the early diagnosis of invasive candidiasis and take advantage of the disease pathophysiology characterized by a 7–10-day delay between exposure to risk factors and infection. We propose a stepwise approach to optimize the accuracy of the CI and other clinical tools (Fig. 3).

Fig. 3
figure 3

Proposed study designs to select high-risk patients who would truly benefit from early empirical antifungal treatment. First, patients should be stratified according to pathophysiological characteristics specific to invasive candidiasis (i.e., immunocompromised individuals, after digestive surgery, other critically ill patients). The three-step approach: In a first optional step, increased intrinsic risk of invasive candidiasis is identified by specific genetic polymorphisms related to innate immunity. In the second step, patients at low objective risk of invasive candidiasis are ruled out by using the high negative predictive values of current risk-assessment strategies (CI, Candida score, peritonitis score, predictive rule). In the third step, empirical antifungal treatment is started early in high-risk patients identified by increased biomarkers values, such as beta-glucan performed only in patients retained by who met criteria outlined in previous steps. A simplified approach restricted to steps two and three may be also of potential interest

Provided that preliminary results can be confirmed in larger cohorts of patients [76], preliminary stratification according to specific genetic polymorphisms could be used as a preliminary step to increase the accuracy of subsequent steps. The second step is to take advantage of the high negative predictive values of the CI, Candida score, and/or predictive rules. This strategy will reduce the currently large proportion of useless antifungal treatments, frequently started for nonobjective emotional reasons. The enhanced intrinsic risk of invasive candidiasis in patients with increased risk identified after this second step may further improve the accuracy of biomarkers, such as mannan/anti-mannan antibodies and/or β-glucan, applied in the third step. This approach should enhance the ability to identify patients who will truly benefit from early antifungal treatment and reduce unnecessary overexposure to antifungal agents in patients with a documented limited risk of infection.

Moreover, the high negative predictive value of the predictive tools (CI, Candida score, and predictive rules) for the further development of an invasive candidiasis may allow one to stop empirical antifungal treatment possibly prescribed before evaluating the patient according to the proposed stepwise approach.

Conclusion

The development of the CI has enabled better appreciation of the dynamics of Candida colonization in patients at high risk of invasive candidiasis. It can still be used, as it is in many institutions, for the early detection of patients at high risk of invasive candidiasis and to guide empirical antifungal treatment. Continued development and execution of clinical trials in this difficult field are important. Until further progress can be achieved with new clinical studies of specific biomarkers involving larger patient cohorts, 20 years after its description, the CI remains one of the best methods of characterizing the dynamics of Candida colonization and identifying patients at very low or increasing risk of invasive candidiasis.