Introduction

Critical care medicine has advanced greatly in the past few decades. Patients with complex medical and surgical disorders are surviving longer due to equally complex medical and surgical interventions. These diseases and interventions often have the “collateral damage” of circumventing normal body defense mechanisms to infection. One such group of infections is invasive candidiasis (IC). The magnitude of this problem in the intensive care setting can be understood if we consider that candidemia is the fourth most common bloodstream infection in the USA [1, 2] (with similar trends around the world [3, 4]) and that 25–50% of cases occur in the intensive care setting [2]. The incidence of IC in the intensive care setting has been studied extensively. Blumberg et al. [5] reported an incidence of 9.8 Candida bloodstream infections per 1,000 admissions in surgical intensive care units (ICUs) in the USA [5]. Leleu et al. [6] reported an incidence of three cases of IC per 1,000 admissions in an ICU in France. In the pediatric population, IC is the second most common bloodstream infection in pediatric and neonatal ICUs in the USA, contributing to 15.8% of all nosocomial infections [7, 8].

IC also has a substantial human and material cost. Attributable mortality has been determined to be as high as 30% [9]. This figure has remained unchanged despite significant advances in prevention and therapy of this disease [10]. Crude estimates of increased costs related to these infections are as high as $44,000 USD per episode [11].

The high morbidity, mortality, and healthcare costs associated with this infection have made it an excellent target for prophylactic, empiric, and preemptive therapy interventions [12–15]. Prevention is, by far, the preferred strategy for dealing with any disease. For prophylaxis to be optimal, a high-risk subpopulation must be found, and a low-cost, effective, and safe measure must be implemented. It is only under these conditions that the risk-benefit and the cost-benefit ratios for prophylaxis will reflect an advantage for the patient. This review will focus on the risk factors and current prevention and treatment strategies for IC in the intensive care setting.

Risk factors for invasive candidiasis

Multiple authors have extensively described the risk factors for IC [1, 5, 8, 9, 16–18]. Table 1 presents a comprehensive list of the risk factors reported in the literature. Most of these factors represent common interventions or conditions in the intensive care setting; unfortunately, research is lacking to determine whether these risk factors have a causal relationship to the diseases or are just associated markers indicating severity of illness and other predisposing conditions. As demonstrated by epidemiological and prophylactic studies, the most important risk factor for IC in the intensive care setting is prolonged length of stay [2, 5, 13, 19]. Most studies have shown that the incidence of IC is relatively low in the first 5–7 days of ICU stay, but by day 7 it dramatically increases, peaking around day 21 [19]. Perhaps one of the most controversial risk factors is colonization by Candida spp. Early studies by Pittet et al. [18] and others [20–22] have shown a proportional increase in the incidence of IC related to the number of body sites that are colonized by Candida spp. Researchers have gone as far as developing a colonization index that is reasonably predictive of risk of IC [18]. Nevertheless, larger scale multicenter studies failed to identify this event as a definite risk factor for the disease [5]. Further research in the multicenter setting is required to clearly define the role of colonization in the pathogenesis and risk of this disease. An interesting observation, though, is that clinical trials that have shown a benefit of fluconazole prophylaxis in the ICU setting often show a benefit in decreasing colonization as well [23].

Table 1 Common risk factors for invasive candidiasis in the critical care setting
Full size table

From defining risk factors to predicting disease

The commonality of the interventions and conditions identified as risk factors for IC renders them practically useless when trying to accurately predict who is actually going to develop a form of IC. This has prompted the development of several strategies or rules to try to predict true risk of disease. Table 2 presents a summary of the most frequently used strategies for risk assessment as derived from primary studies to define such strategies or from clinical trials of prophylaxis.

Table 2 Risk assessment strategies for invasive candidiasis (IC) in the critical care setting
Full size table

As seen in Table 2, the early attempts to establish risk were based on the presence of colonization, but as stated above, the jury is still out on the relevance of this risk factor, although many centers use the colonization index as a routine way to assess risk of IC and before treating patients preemptively or empirically. The next set of risk-assessment strategies comes from clinical trials of ICU prophylaxis. These clinical trials usually rely on increased length of stay plus the presence of one or more risk factors or markers of severity of illness and prolonged stay, such as intubation [23], for patient enrollment. More recent work has focused on scoring systems or risk prediction rules that have been developed retrospectively and have yet to be proven prospectively in multicenter settings. Notable examples of these are the prediction rules by Paphitou et al. [24] and Ostrosky-Zeichner et al. [25] and the scoring system developed by DuPont et al. [26]. Several scoring systems are also being validated at this time, and the findings are awaiting publication.

Paphitou et al. [24] performed a retrospective chart review of surgical intensive care patients who stayed ≥4 days over a 1-year period in a surgical/trauma ICU in the USA. Patients that had any combination of diabetes mellitus, new-onset hemodialysis, use of parenteral nutrition, and broad-spectrum antibiotics had a rate of IC of 16%, while the rate in patients who did not have a combination of factors was 5%. Unfortunately, this definition of risk applied to an overwhelming 52% of patients, but on the other hand, it captured 78% of all cases of IC in that unit. A drawback of this study is that, in an effort to acknowledge clinically relevant scenarios, the definitions of IC were broad, including empiric antifungal use as a definition of possible IC, thus decreasing the validity of diagnosis.

On the basis of the work of Paphitou et al. [24], Ostrosky-Zeichner et al. [25] created a prediction rule for IC in a retrospective multicenter setting (12 units in the USA and Brazil). The rule requires that a combination of one “major” and two “minor” criteria be met, with the criteria reflecting common risk factors for patients that stay in the unit >3 days and are expected to stay for an additional 2 days. This rule applied to roughly 10% of patients who stay in the unit more the 4 days, and ∼10% of patients to whom this rule is applied will develop proven or probable IC. This rule is now being validated prospectively in a multicenter clinical trial of Candida prophylaxis in the intensive care setting.

DuPont et al. [26] performed a retrospective systematic review of surgical intensive care patients in France with prospective follow-up. A scoring system was created using the following risk factors: female gender, upper gastrointestinal origin of peritonitis, cardiovascular failure, and use of antibiotics. A grade C score (presence of 3 qualifiers) was associated with a sensitivity of 84% and specificity of 50% for the detection of yeast in the peritoneal fluid of patients with peritonitis. The drawbacks of this study are its single-center setting and its use of a disease definition of limited scope.

Clinical trials on prophylaxis

There are three major clinical trials on prophylaxis that have been conducted to date, all of which studied fluconazole as the prophylactic agent. Eggiman et al. [27] conducted a prospective, randomized, placebo-controlled study of fluconazole for prophylaxis of intra-abdominal candidiasis in surgical patients from two university hospitals in Switzerland. The study enrolled 43 patients with recent complicated surgeries (recurrent perforations or anastomotic leakage). The primary endpoint was development of intra-abdominal candidiasis, with the secondary endpoints being development of other Candida infections and Candida colonization. The frequency of Candida peritonitis was reduced from 35% to 4% in the fluconazole group. Regarding colonization, the authors showed that colonization occurred in 62% of the placebo group versus 15% of the fluconazole group. All of these differences were statistically significant.

Pelz et al. [19] studied a surgical unit in the USA. This was a prospective, randomized, placebo-controlled trial of fluconazole in critically ill surgical patients. The risk factor used by the authors was a potential length of stay of >3 days. The main endpoint was time to development of fungal infection in the intent-to-treat population. The definitions of fungal infection were broad yet compatible with the currently used EORTC/MSG criteria for diagnosis of fungal infection [28]. In 260 enrolled patients, the overall incidence of proven fungal infections was 15.4% versus 8.5% in the fluconazole group. At day 14 of ICU stay, the probability of infection as determined by Kaplan-Meier estimates was 0.4 overall versus 0.13 in the fluconazole group. In the multivariate analysis, significant predictors of IC included APACHE II score, days to first dose of the prophylactic drug, fungal colonization at enrollment, and parenteral nutrition, thus confirming once more the relevance of these risk factors. Interestingly, during follow-up to this study, the authors continued to show the value of their prophylactic measure without having a shift to non-albicans Candida species, but they failed to show any differences in mortality [29].

Garbino et al. [23] studied two units (1 medical, 1 surgical) in a university hospital in Switzerland. The study was a prospective, randomized, placebo-controlled trial of fluconazole in 204 patients. The main enrollment criterion was mechanical ventilation for at least 48 h and an expected duration of stay in the unit of at least 72 h. The main outcome was incidence of Candida infection using non-standardized definitions. The incidence of IC was 16% overall versus 5.8% in the fluconazole group. As in the other trials that have studied fluconazole, Candida colonization was significantly decreased over time in the patients who received fluconazole. Noteworthy, too, is that the authors also looked at mortality in this study, but again, no difference between the two groups could be shown at the end of the study.

These clinical trials have shown that fluconazole prophylaxis may be considered effective in reducing the incidence of Candida infections in selected critically ill patients, both in the medical and surgical critical care settings. They also confirm the major risk factors for IC. The major drawbacks of these studies are the lack of standardized disease definitions and the geographical limitation to one or two hospitals. Large multicenter clinical trials are under way in the USA and Europe. This “next generation” of trials should be answering the questions needed for a paradigm change [13, 14] and thus settle once and for all the usefulness of this measure. Future studies should probably continue to measure the effect of prophylaxis on gross and attributable mortality, but they should continue to use the measurement as a secondary endpoint. With regard to the agents of choice, one must consider that use of the echinocandins may be justified, particularly since these agents are safe and highly effective [30–33] and since severe changes in the epidemiology of Candida spp. have been observed, with a higher incidence of non-albicans species showing potential fluconazole resistance [3, 16, 34, 35].

Prophylaxis versus preemptive therapy: the need for more sensitive diagnostic markers

Perhaps an interesting discussion for the future will be, “Which strategy is more useful: universal/targeted prophylaxis or preemptive therapy of infections documented before they produce clinical signs and symptoms or positive cultures?” Current methods for culture of Candida are less than optimal. Blood cultures are positive in only 50% of the patients with IC [36]. The reason for asking the above question stems from the recent development of surrogate markers for IC. Surrogate markers, or non-culture-based methods, for early detection of Candida are being actively developed in this setting [37, 38]. Antigen and antibody detection systems have been disappointing to date [39, 40], and measurement of metabolites such as d-arabinitol has been impractical or complicated [41], but there are many new techniques that measure fungal genetic products by PCR or fungal cell wall components (such as β-glucan) by a variety of assays [37, 42, 43]. The use of all of these methods, tied to rigorous economic analyses, will define which strategy will have the best cost-benefit and risk-benefit ratios.

Therapy of invasive candidiasis in the intensive care setting

Treatment guidelines drafted by experts in the USA for the management of IC have been published [44] and recently updated [45]. These guidelines mention the following drugs as appropriate first-line agents for these diseases: amphotericin B, fluconazole, and caspofungin. The efficacy of these agents has been investigated in well-documented trials in the literature [46, 47]. The choice between these agents will depend on many factors that are particularly relevant in the intensive care setting. The first one is knowledge of the local epidemiology of the unit. Fluconazole may be a poor choice for empirical therapy in units with a high incidence of Candida glabrata and Candida krusei, due to the acquired or intrinsic resistance that these species exhibit [44, 48], while it may be a perfectly appropriate choice if Candida albicans is still the predominant organism. Another factor is the criticality of the patient. When a patient is critically ill, one may want to reduce the chance that the patient will be treated with a drug for which there is potential resistance, or one may prefer a drug with the theoretical advantage of being rapidly fungicidal; in such situations, one would choose amphotericin B (deoxycholate or its lipid formulations) or caspofungin rather than fluconazole. If, however, the patient is relatively stable, one may choose to use fluconazole upfront and thus avoid the use of the highly nephrotoxic amphotericin B deoxycholate or the costly lipid-based amphotericin B compounds and caspofungin. Finally, another factor to consider is the side effect profile of the drugs and the organic reserve of the patient. When treating a patient with renal failure, one would clearly stay away from amphotericin B deoxycholate and choose instead caspofungin, reduced-dose fluconazole, or a lipid-based amphotericin B compound.

Regarding resistance and susceptibility testing, one may want to concentrate on identifying the isolate to the species level and reserve susceptibility testing for special situations [45, 49–51]. The species of the organism is generally thought to reliably predict the organism’s susceptibility to the different antifungal agents. Candida krusei should be considered uniformly resistant to azoles, while Candida glabrata may be resistant or require high doses of azoles for successful treatment. Candida lusitaniae should be considered intrinsically resistant to amphotericin B, and MICs of echinocandins for Candida parapsilosis are typically higher, although there has not been any clinical evidence to suggest that this in fact correlates with resistance [47]. Susceptibility testing should be generally reserved for patients in whom therapy has failed and in whom therapeutic options are limited [45, 49–51].

General principles of therapy include removing all compromised lines, devices, and implants, if possible, since the removal of these foreign objects actually correlates with better patient outcomes [52, 53]. Biofilms appear to play a major role in the persistence and proliferation of IC [53, 54]. The echinocandins may have an advantage in this setting, since experimental data has shown they have penetration and action in Candida biofilms [55]. Nevertheless, one should consider that long-term catheters are at low risk for being the source of infection [52, 53] and that neutropenic patients may have other sources of candidemia, such as gut translocation [56]. The recommended duration of treatment is 14 days since the first negative culture if candidemia is present (and disseminated disease is excluded), or until clinical, microbiological, or radiological resolution if other forms of IC are diagnosed. It is recommended that all candidemic patients have a dilated eye exam to exclude endophthalmitis and disseminated disease, since this has obvious implications for the duration of therapy [45, 57].

Is combination therapy useful for these diseases? Early experience and the guidelines [44, 45] recommend the combination of amphotericin B and flucytosine for particularly severe disease, mainly endocarditis or severe sepsis. Nevertheless, a recent multicenter, randomized, double-blind, placebo-controlled trial of fluconazole versus fluconazole plus amphotericin B [58] showed (despite an imbalance in APACHE II scores between the two groups) that although the combination had no advantages over monotherapy in most of the endpoints, blood cultures did clear faster in the group that received combination therapy, thus offering a theoretical advantage. Another important finding of this study is that no antagonism between the two drug classes was demonstrated in clinical practice.

Perhaps one of the biggest issues is that of empiric therapy in a febrile high-risk patient with negative cultures. The typical patient is critically ill, on broad-spectrum antibiotics, has some Candida colonization (urine and sputum), and is febrile or hemodynamically unstable. Although there is no published case-controlled study or clear guidance for this situation, it is this author’s personal bias to empirically treat patients that have multiple risk factors with a drug selected according to the criteria mentioned above. As discussed, future studies should better define the value of prophylactic, preemptive, and empiric therapy in the critical care setting.