Bench-to-bedside review: Therapeutic management of invasive candidiasis in the intensive care unit

Candida is one of the most frequent pathogens in bloodstream infections, and is associated with significant morbidity and mortality. The epidemiology of species responsible for invasive candidiasis, both at local and worldwide levels, has been changing - shifting from Candida albicans to non-albicans species, which can be resistant to fluconazole (Candida krusei and Candida glabrata) or difficult to eradicate because of biofilm production (Candida parapsilosis). Numerous intensive care unit patients have multiple risk factors for developing this infection, which include prolonged hospitalisation, use of broad-spectrum antibiotics, presence of intravascular catheters, parenteral nutrition, high Acute Physiology and Chronic Health Evaluation score, and so forth. Moreover, delaying the specific therapy was shown to further increase morbidity and mortality. To minimise the impact of this infection, several management strategies have been developed - prophylaxis, empirical therapy, pre-emptive therapy and culture-based treatment. Compared with prophylaxis, empirical and pre-emptive approaches allow one to reduce the exposure to antifungals by targeting only the patients at high risk of candidemia, without delaying therapy until the moment blood Candida is identified in blood cultures. The agents recommended for initial treatment of candidemia in critically ill patients include echinocandins and lipid formulation of amphotericin B.

aff ect up to about 10% of all admitted patients [18,19]. Additionally, Candida species account for approximately 3% of all surgery-related peritoneal infections, both community-acquired infections and nosocomial infections [20]. A recent French study on invasive candidiasis in the ICU revealed that approximately one-third of patients presented each of the clinical forms: isolated candidemia, invasive candidiasis with candidemia and invasive candidiasis only [1].
Th e epidemiology of Candida infections -both on a worldwide scale and, more importantly, on a local levelhas signifi cant implications for management of this infection. In particular, given the well-known, albeit not universal, diff erences in antifungal susceptibility among diff erent Candida species, the choice of the most appropriate empirical treatment can be successfully based on epidemiological data regarding the frequency of Candida parapsilosis and fl uconazole-resistant species in a given centre.
During the past two decades, most hospitals have reported a progressive shift in the species of Candida. In the past, almost all isolates responsible for bloodstream infections were Candida albicans, whereas in recent years a growing proportion of episodes of candidemia have been caused by Candida species other than C. albicans [21][22][23][24][25]. Even though, C. albicans remains the predominant strain in most countries [26,27], even among critically ill patients [1,7,21,28,29], non-albicans species are increasingly common -some ICUs have reported recently that non-albicans species are responsible for over 50% of candidemia episodes in adult critically ill patients [24,30]. Th e most common nonalbicans species are C. parapsilosis or Candida glabrata, followed by Candida tropicalis and Candida krusei [1,24,[31][32][33]. Rare Candida species reported to cause candidemia include Candida lusitaniae, Candida guilliermondii and Candida rugosa [15,31].
Numerous studies have tried to fi nd reasons for this shift, and several risk factors have been associated with candidemia due to diff erent species [25,34]. It is understandable that the widespread use of fl uconazole can predispose patients to the development of infections due to species that are resistant to azoles, either intrinsically fully resistant such as C. krusei or in dose-dependent fashion such as C. glabrata. Indeed, previous use of fl uconazole has been found to be a risk factor for the presence of non-albicans fungaemia [24,25,35] even though some studies did not fi nd this association [23]. Specifi c risk factors for candidemia due to other nonalbicans strains have also been reported, such as for example the presence of in-dwelling devices, hyperalimentation and being a neonate for C. parapsilosis [31]. Th e specifi c risk factors associated with diff erent Candida species are outlined in Table 1.
Even though the overall rise in incidence of nonalbicans strains is alarming, from the clinical point of view there are important diff erences among the diff erent species. Specifi cally, the main diff erence between C. albicans and C. kusei or C. glabrata is the resistance to the most frequently used antifungal (that is, fl uconazole). Th e diff erences in susceptibility to various antifungals are partially predictable and are reported in Table 2. Species identifi cation and knowledge of the local epidemiology of Candida strains causing candidemia are therefore of the utmost importance for guiding appropriate empirical therapy. On the contrary, in vitro susceptibility testing of clinical isolates of Candida proves extremely valuable for guiding therapy in patients who have received prior antifungal treatment or who are not responding to empirical therapy.

Risk factors for invasive candidiasis in the ICU and predictive scores
Although invasive Candida infections can aff ect any hospitalised patient, they are more common and have unique attributes in certain populations, including patients with cancer, haematological malignancy or other immunosuppression. Th e predominant source of invasive Candida infection is endogenous, from superfi cial mucosal and cutaneous proliferation to haematogenous dissemination [36]. Rare cases of exogenous transmission have been described due to contaminated solutions and materials or transmission from healthcare workers to patients and from patients to patients [37,38].
Th e suppression of the normal bacterial fl ora in the gastrointestinal tract by broad-spectrum antibiotic Table 1. Particular risk factors associated with candidemia due to diff erent Candida species therapy also allows the yeast to proliferate, both in neutro penic patients with haematological malignancies [39] and in non-neutropenic patients [40], and long-term and high-density colonisation has been shown to lead to candidemia. Numerous conditions frequent in ICU patients -such as parenteral nutrition, intravascular catheters, trauma, hypotension, therapy with steroids or cyclosporine, and ischaemia and reperfusion -may damage the integrity of skin or the gastrointestinal mucosa with penetration by the yeast, potentially leading to systemic infection. Th e factors predisposing critically ill patients to candidemia are presented in Table 3, with the presence of vascular catheters or disruption of the gut or skin barrier among the most important.
Important eff orts are focused on identifying critically ill patients at high risk of developing candidemia in order to apply the most effi cacious management strategy and avoid high mortality. Risk prediction scores have thus been developed and diff erent parameters combined to predict which patients would develop candidiasis. In parti cular, the score by Leon and colleagues included parenteral nutrition (1 point), surgery (1 point), multifocal colonisation (1 point), and severe sepsis (2 points); subjects with score >2.5 were almost eight times more likely to develop candidiasis than those with score <2.5 [41]. Th e other score by Ostrosky-Zeichner and colleagues found that the combination of the systemic antibiotic treatment or central venous catheter and two or more of fi ve other variables (parenteral nutrition, dialysis, major surgery, pancreatitis, treatment with steroids or other immuno suppressive agents) had, in their population, positive and negative predictive values of 10% and 97%, respectively [42]. Additionally, Dupont and colleagues published in 2003 a predictive score for peritoneum Candida infection in the ICU: the presence of three out of four factors (female gender, upper gastrointestinal tract origin of peritonitis, intra operative cardiovascular failure and previous antibiotic therapy) had positive and negative predictive values of 67% and 72%, respectively [43].
Th anks to such scores specifi cally developed and then validated in ICU patients, the risk of invasive candidiasis can be estimated because the presence of the abovementioned risk factors is directly related to the percentage probability of developing invasive candidiasis, allowing one to judge whether risk of candidemia warrants any therapeutic measures.

Diagnosis of candidemia
Blood cultures remain the mainstay for diagnosing candidemia, but the sensitivity reported frequently is not optimal. Moreover, the time from blood sample collection and the microbiological response of growing yeast is often lengthy. Furthermore, several more days are required for species identifi cation and susceptibility testing. New methods for diagnosis of invasive Candida infection have therefore been investigated, including  serological markers (mannan and β-d-glucan) and realtime PCR; however, only the use of β-d-glucan has been included in the 2008 Infectious Diseases Society of America guidelines for diagnosing invasive fungal disease [44]. Th e use of β-d-glucan is currently being investigated in ICU populations. Even though the results seem promising, no large prospective studies have been performed and the main problems for the use of β-dglucan remain its high cost and high rate of false positive results (mostly due to concomitant bacterial bloodstream infections and intensive care measures such as haemo fi ltration, albumin or immunoglobulin use) [45]. Traditional culture from sterile sites other than the bloodstream (for example, the peritoneum) are useful for diagnosis of invasive candidiasis. For specifi c details on the diagnosis of invasive candidiasis in the ICU, a recent review is available [46].

Management of candidemia in the ICU
As far as management of candidemia in the ICU is concerned, there is no single strategy that can be considered the most appropriate. In fact, diff erent approaches can be chosen and can be judged as the best for a given clinical situation. In particular, four management options are available: prophylaxis, empirical therapy, pre-emptive therapy and treatment of a cultureproven infection. So how is the best strategy chosen?
Th e knowledge of epidemiological data, the abovementioned risk factors and, fi rst of all, the analysis of local epidemiology of candidemia in a singular ICU allow one to determine whether a patient is at low, modest or high risk of developing this infection. Consequently, a choice between the most appropriate management strategies can be made -the patients with low or modest risk of infection can be monitored less frequently for Candida colonisation, while high-risk subjects may benefi t from immediate diagnostic procedures (cultures of both sterile and nonsterile sites, testing for serological markers) and empirical antifungal therapy. In the case of negative results of testing for yeasts, antifungal prophylaxis might be considered. Naturally, knowing the most frequent species and susceptibility patterns of Candida isolated in a single ICU is the basis for choosing an adequate antifungal agent (Table 4).

Prophylaxis in the ICU
Prophylaxis -defi ned as administration of an antifungal agent to a patient with no evidence of infection -has been evaluated in several studies and meta-analyses [27,[47][48][49][50], and its main advantage is a possible reduction in the rate of candidemia.
Since morbidity and mortality rates in patients with systemic fungal infections are exceedingly high, the use of an eff ective antifungal prophylaxis in selected highrisk patients is very attractive and might be an option in this select population. Th e strategy of antifungal prophylaxis is now well established in patients with persistent neutropenia after treatment for haematological malignan cies or after bone marrow transplantation, but routine use of antifungal prophylaxis in the general ICU setting is discouraged [51][52][53]. Nonetheless, the implementation of targeted anti fungal prophylaxis has been shown to be eff ective in certain ICU settings [27,47], and three randomised placebo-controlled trails reported a clear >50% decrease in the incidence of Candida infections with fl uconazole prophylaxis [54][55][56]. Moreover, two meta-analyses con fi rmed that prophylactic fl uconazole administration in ICU patients reduced the rate of Candida infection, but no clear survival advantage was observed [49,50]. In the meta-analysis by Playford and colleagues, however, only when the studies on prophylaxis with both fl uconazole and ketonazole were considered was the total mortality found reduced by approximately 25% and the incidence of fungal infection by 50% [50]. Th e meta-analysis by Cruciani and colleagues also reported, along with a relative risk reduction in candidemia, a decrease in overall mortality with antifungal prophylaxis with various agents [48].
On the other hand, the disadvantages of fl uconazole prophylaxis include possible toxicity and profound infl uence on local epidemiology with the emergence of fl uconazole-resistant isolates [57]. From expert opinion expressed in reviews and guidelines [44,58], therefore, antifungal prophylaxis might be warranted only for ICUs with a high rate of invasive candidiasis, as compared with the normal rates of 1 to 2%, particularly for selected patients who are at highest risk (>10%) [42]. Th e approach of limiting prophylaxis to a subgroup of patients with the highest risk of candidemia may help to limit the quantity of antifungals used and delay the emergence of infections due to fl uconazole-resistant Candida strains seen in immunocompromised patients. In fact, this approach is supported by the recent Infectious Diseases Society of America guidelines that recommend fl uconazole prophylaxis at a dose of 400 mg (6 mg/kg) daily for high-risk adult patients hospitalised in ICUs that have a high incidence of invasive candidiasis [44].

Empirical therapy
Empirical treatment is defi ned as administration of antifungals in the presence of persistent and refractory fever in patients who are at high risk for fungal infection. Th is strategy was developed almost three decades ago for neutropenic patients, when it became evident that the lack of sensitivity of microbiological and clinical fi ndings resulted in delayed diagnosis and increased morbidity and mortality.
Even though the fi rst studies on empirical therapy were underpowered, the treatment being used in diff erent clinical settings and numerous antifungals are being registered and recommended for empirical treatment of invasive candidiasis, both in neutropenic patients and in non-neutropenic patients [44]. All these eff orts are aimed at reducing morbidity and mortality by starting treatment as early as possible, given the evidence that a delay in antifungal prescription increases mortality rates significantly in candidemia [13,14]. In the ICU, however, where numer ous patients have diff erent risk factors for fungal infections, the routine use of empirical therapy in cases of persistent fever may result in signifi cant overtreatment. Th e strategy of a pre-emptive approach therefore appears promising. In fact, US guidelines recommend that such an approach (although they continue to call it empirical treatment) should be considered for critically ill patients with risk factors for invasive candidiasis and no other known cause of fever, based on clinical assessment of risk factors, serologic markers for invasive candidiasis and/or culture data from nonsterile sites [44] (Table 4).

Pre-emptive therapy
Th e main concept of a pre-emptive strategy is to better identify patients at high risk for developing candidemia. Th e overall use of antifungals in the ICU can therefore be reduced, without delaying therapy in patients who need it. Th e recent availability of more sensitive and specifi c clinical and laboratory tools allows for better identifi cation of high-risk patients, and this approach has been used success fully [59]. Th e question arises, however, of how to defi ne a patient at high risk for developing candidemia. No clear predictive rule exists, but the two score systems described above for ICU patients can be of some help. In brief, multifocal colonisation by Candida and/or the presence of well-described factors outlined in Table 3 make the patient a suitable candidate for empirical therapy if any signs or symptoms of infection compare.
In particular, the effi cacy of a pre-emptive strategy in ICU patients has been recently established in a singleinstitution study in which the use of fl uconazole in patients with corrected colonisation index ≥0.5, des cribed previously by Pittet and colleagues [40], has signifi cantly decreased the incidence of invasive candidiasis [60]. More over, surrogate markers of invasive fungal infections have been studied extensively. In particular, β-d-glucan is a component of the cell wall of Candida and other fungi and has been investigated as a serological marker for fungal infections, including candidemia [61]. Even though false positive results have been reported and its routine use in the ICU requires further validation, persistently high serum levels of β-dglucan in ICU patients were found indicative of fungal disease [45].
A pre-emptive approach in critically ill patients might therefore be defi ned as starting antifungals when the following conditions are satisfi ed: the presence of long ICU stay (>96 hours), and broad-spectrum antibiotic therapy; the presence of any other risk factors, such as severe sepsis, gastrointestinal surgery or parenteral nutrition; plus microbiological evidence of Candida infection, including multifocal colonisation or a positive result for serum β-d-glucan. Th e proposed approach is shown in Figure 1 and, as with any new strategy, will warrant valida tion in prospective trails. One of the main advantages of such an approach is limiting the use of antifungals in low-risk patients, while starting treatment for candidemia without delay when symptoms appear in patients at high risk for this infection. Th e benefi t of early therapy, in terms of morbidity and mortality, can thus be obtained, while overtreatment can be avoided (Figure 2).

Treatment of a culture-documented candidemia
For ICU patients with low/medium risk of developing candidemia, blood cultures should be performed if a clinical suspicion of systemic infection is present, even in the absence of fever. Numerous blood cultures, both from a central venous catheter and a peripheral line, remain the cornerstone for diagnosis of candidemia. As any delay before administering primary therapy can lead to a noticeable increase in mortality, antifungals should be prescribed as soon as there is growth of yeasts in blood samples.
Th e choice of antifungal for an unknown Candida species should be based on the knowledge of local  [44]. Th e antifungal treatment might be modifi ed according to the results of susceptibility testing, and de-escalation to fl uconazole has been successful for stable patients with susceptible isolates [44].

Other aspects of treating candidemia in the ICU
Once the initial therapy for candidemia is started, several clinical issues remain open. Firstly, the effi cacy of the treatment should be assessed by the documentation of blood cultures returning sterile. Moreover, the date of the fi rst negative blood culture is important, because the recommended length of treatment is 14 days after the documented clearance of Candida from the bloodstream and resolution of symptoms attributable to candidemia. Secondly, the antifungal chosen empirically can be changed based on the results of species determination or susceptibility testing. For stable patients with C. albicans or other fl uconazole-susceptible strains, fl uconazole is therefore the drug of choice. Importantly, fl uconazole is the preferred treatment for C. parapsilosis, since resistance to echinocandins has been reported [62].
Th irdly, patients who improved clinically and who cleared Candida from the bloodstream might be suitable for step-down oral therapy to complete the course of 14 days. Th e available oral antifungals are fl uconazole, itraconazole, voriconazole and posaconazole. Fluconazole is an obvious choice for susceptible species, while voriconazole can be indicated as step-down therapy for C. krusei or voriconazole-susceptible C. glabrata.
Additionally, ophthalmologic fundus examination is warranted in all the patients to exclude disseminated endocular infection, and endocarditis should be excluded in case of persistently positive blood cultures, known valve pathology or any other sign or symptom suggestive of endocardial involvement. In both cases, the duration of treatment is much longer (>4 weeks and up to lifelong suppressive therapy) and is described in detail elsewhere [44].
Last but not least, intravenous catheter removal is strongly recommended for patients with candidemia. Guidelines both on management of candidiasis and on management of catheter-related bloodstream infection state clearly that catheters should be removed, even though grade II and grade III of such statements indicate that there are no data from properly randomised, controlled trials [44,63,64]. Interestingly, a recent study of 842 adults included in candidemia trials did not fi nd on multivariate analysis any benefi t from early catheter removal; the expert guidelines remain the best synthesis of all available data, however, and removing the catheter should thus be attempted in all ICU patients with candidemia [64].
Biofi lm production is a well-documented phenomenon for Candida species that signifi cantly contributes to Candida pathogenicity in catheter-related bloodstream infections, resulting in recurrent or persistent infections and biofi lm-mediated antifungal resistance leading to treatment failure [65]. Moreover, the mortality in patients with invasive infections due to biofi lm-producing Candida species has been reported signifi cantly higher [66]. Th e activity of antifungals against biofi lm therefore has important clinical implications and is known to vary among diff erent agents. In particular, fl uconazole and azoles -which are static against Candida -are also not active against sessile forms, while echinocandins and amphotericin B off er both bactericidal activity and good penetration into a biofi lm formed on vascular devices. However, a study performed on 43 Candida speciesincluding 12 C. albicans, 12 C. parapsilosis, 10 C. tropicalis and nine C. glabrata isolates -found that the activity of caspofungin and micafungin against a biofi lm of C. parapsilosis and C. tropicalis was signifi cantly lower than that of amphotericin B [67]. Table 5 outlines the susceptibility of diff erent Candida species to two antifungals that are active against biofi lm-producing strains.

Antifungal agents
In recent years, numerous new antifungal drugs have been developed, studied and approved for various indications, and almost all of these new drugs are licensed to treat candidemia in diff erent patient populations. Th e most appropriate antifungal drug can be chosen from the three main groups: the polyenes (amphotericin B deoxycholate, lipid complex, liposomal); the azoles (fl uconazole, voriconazole, posaconazole, itraconazole, ravuconazole); and the echinocandins (caspofungin, micafungin, anidulafungin).
Most of the studies on effi cacy in candidemia have not shown signifi cant diff erences between various agents. Th e diff erences in drug-related toxicity are signifi cant, however, and the possibility of drug-drug interactionsso important in critically ill patients that receive numerous medications -varies signifi cantly among the single agents. Th e choice of the best antifungal therefore still poses a challenge for a clinician. Th e detailed description of various agents used for treating candidemia is beyond the scope of the present review, but the dosing of the main antifungals is reported in Table 6. Moreover, given that echinocandins are the most recently introduced class of antifungals and general recommendations do not usually specify which of them should be used, Table 7 outlines the diff erences in indication, dosing, and so forth, for three echinocandin compounds. Considering that many ICU patients have other signifi cant comorbidities, data on the treatment with various antifungals in the case of renal or hepatic insuffi ciency are reported in Table 8.

Management of candidemia in the neonatal ICU
Th e incidence of candidemia in the neonatal ICU has been increasing, mostly due to the fact that more lowbirth-weight and very-low-birth-weight newborns sur vive longer thanks to advances in medical technology. Th ese  newborns are more likely to develop infectious complications, and candidemia is one of the most frequent nosocomial bloodstream infections in this population. Th e reported risk factors for candidemia in neonates and adults are similar, and include central venous catheters and arterial lines, parenteral nutrition, mechanical ventilation, and the extended use of antibiotics. Unlike in the adult ICU, C. albicans remains the most common isolate in the neonatal ICU -although non-albicans species such as C. parapsilosis and C. tropicalis are increasingly common [68,69]. Fortu nately, these species are susceptible to fl uconazole. Th e recent Infectious Diseases Society of America guidelines on management of Candida infections off er recommendations for paediatric patients. In particular, the following treatments are regarded as fi rst-line therapy for neonatal candidiasis: amphotericin B deoxycholate, or liposomal amphotericin B if urinary tract involvement is excluded, and fl uconazole. Th e guidelines also state that echinocandins should be used with caution and are  generally limited to situations in which resistance or toxicity precludes the use of fl uconazole or amphotericin B. Dosing of antifungals in paediatric patients is outlined in Table 9.
Additionally for neonates, a lumbar puncture and a dilated retinal examination -preferably by an ophthalmologist -are recommended in those with sterile body fl uid and/or urine cultures positive for Candida, and removal of the intravascular catheter is strongly recommended. Finally, in nurseries with high rates of invasive candidiasis, fl uconazole prophylaxis may be considered in neonates with birth weight <1,000 g.

Conclusions
Candida is one of the most common causes of nosocomial bloodstream infection. Morbidity and mortality associated with candidemia are signifi cant and the epidemiology of species has been changing, at both local and worldwide levels. Even though numerous risk factors for invasive Candida infection have been reported and several antifungals are widely available, the optimal management of candidemia remains a challenge. Th e agents recommended for initial treatment of candidemia in critically ill patients include echinocandins and lipid formulation of amphotericin B, but the choice between prophylactic, empirical and pre-emptive therapy is crucial. Compared with prophylaxis, empirical and preemptive approaches allow the clinician to reduce exposure to antifungals by targeting only patients at high risk of candidemia, without delaying therapy until yeast is identifi ed in blood cultures. A pre-emptive strategy is based on the presence of numerous risk factors, together with micro biological documentation for the presence of Candida, such as multifocal colonisation or positive serum β-d-glucan. Further prospective studies are  warranted to confi rm the benefi ts from routine use of