Diagnosis and management of invasive candidiasis in the ICU: an updated approach to an old enemy

In the USA, Candida species are responsible for 8–10 % of bloodstream infections (BSI) and are the fourth most common bloodstream pathogen. In Europe, Candida species only account for 2–3 % of BSI and are ranked as the 6th–10th most frequent pathogen [3]. Approximately 30–35 % of all episodes of candidemia occur in ICU patients [4]. In the Extended Prevalence of Infection in Intensive Care (EPIC II) point prevalence study [5], Candida was isolated in 17 % of infected culture-positive patients. Several studies have reported increasing incidence rates of candidemia in ICUs over recent years [6, 7], although population-based studies have suggested reduced incidence rates possibly related to improved infection control practices in high-risk patients [8, 9].

Invasive candidiasis is a highly lethal infection associated with mortality rates between 40 and 60 % [6, 7, 10, 11]. The five most common Candida species are Candida albicans, Candida glabrata, Candida tropicalis, Candida parapsilosis, and Candida krusei [12]. C. albicans was previously the predominant species isolated in patients with invasive candidiasis, accounting for 65–70 % of the total number of Candida isolates. However, the epidemiology has changed in recent years, with non-albicans species now responsible for about half of the cases in some centers [10, 13, 14]. C. parapsilosis tends to be more frequent in southern Europe, Australia, and Latin America than elsewhere [10, 15‐17], and C. glabrata is more frequent in the older population than in middle-aged adults [7, 15]. Non-albicans species often have reduced susceptibilities or even intrinsic resistance to azoles or echinocandins [17]. These epidemiological trends underscore the need to perform large longitudinal studies to obtain data that can be used to guide empirical antifungal therapy.

Prompt and accurate diagnosis of invasive fungal infection is crucial so that appropriate antifungal agents can be started rapidly. However, early diagnosis is not always easy. Microscopic examination is rapid and can be helpful but a negative result does not exclude infection [18, 19]. Blood cultures are positive in only 50–70 % of cases of Candida BSI [20]. Furthermore, it can take several days before Candida is identified at the species level and antifungal susceptibility data are available. Moreover, blood cultures are rarely positive in patients with deep-seated candidiasis [20]. In such patients, cultures of infected tissues can be performed but have their own limitations, including the need for invasive surgical procedures and poor sensitivity [20].

Recently, nonculture-based diagnostic tests have been developed for detection in blood of components of the fungal cell wall (such as mannan and β-d-glucan (BDG)) by immunoassays, of DNA by PCR, and of antibodies by serology. The performance of BDG assays in invasive fungal infections has been assessed in three systematic reviews. In the subgroup of patients with proven infection, the pooled sensitivities and specificities of BDG were 79.1, 87.7, and 78 %, respectively [21‐23]. Importantly, BDG assays are not species specific so further tests are needed to identify the specific fungus. In patients with suspected invasive candidiasis, the presence of elevated circulating levels of Candida mannan antigen in the blood had a sensitivity of 58 % and a specificity of 93 % [24]. In the same group of patients, the detection of anti-mannan antibodies had a sensitivity of 59 % and a specificity of 83 %. The sensitivity increased to 83 % when the mannan and anti-mannan assays were combined, but the specificity remained similar at 86 %. PCRs for fungi continue to be challenging for technical reasons and there are no commercially available tests at present. In a review of more than 50 standard, nested, or real-time PCRs [25], the pooled sensitivity and specificity were 95 and 92 %, respectively, for invasive candidiasis.

Combining several markers may prove more useful. A technology combining PCR and nanoparticle-based hybridization, the T2 magnetic resonance-based biosensing technology platform, can detect as little as 1 colony-forming unit/ml of the five most common Candida species [26] and was shown to have an overall specificity of 99.4 % for invasive candidemia [27]. This test is expensive to perform, however, and although costs may be balanced by more accurate earlier diagnosis with reduced administration of antifungal agents and reduced mortality rates [28], further studies are needed to validate its use and assess its cost–benefit ratio.

Numerous risk factors for invasive Candida infections have been identified, including higher Acute Physiology and Chronic Health Evaluation II scores, diabetes mellitus, renal insufficiency, surgery (especially abdominal surgery), pancreatitis, the use of broad-spectrum antibiotics, parenteral nutrition, hemodialysis, mechanical ventilation, the presence of central vascular catheters, and therapy with immunosuppressive agents [29‐34]. The development of invasive Candida infections is often preceded by extensive colonization of the skin or of the mucus membranes of the gastrointestinal and urogenital tracts, and the degree of colonization, assessed using the Colonization Index, has been shown to be an independent risk factor for development of candidiasis [35, 36].

Several prediction rules and scores based on clinical, laboratory, and microbiological parameters have been proposed to help clinicians identify patients at high risk of developing invasive fungal infections [32, 33, 35, 37‐41]. Ostrosky-Zeichner et al. [40] proposed a prediction rule characterized by a very high negative predictive value (0.97) and including the following parameters: use of systemic antibiotic therapy, total parenteral nutrition, dialysis, steroids or immunosuppressive agents, major surgery or pancreatitis, and the presence of a central venous catheter. The Candida score, an easy-to-use assessment system proposed by Leon et al. [32, 33], integrates four risk factors (total parenteral nutrition, surgery, multifocal Candida colonization, and severe sepsis) and also has a high negative predictive value (0.98) to rule out invasive candidiasis [33]. However, colonization surveillance cultures are expensive and their precise value remains unclear [42].

More recently, biomarkers of fungal infection have been suggested to assist in decisions to start or stop antifungals. BDG was superior to the Colonization Index or the Candida score for the prediction of intraabdominal candidiasis in high-risk surgical and ICU patients, with a cutoff value of 80 pg/ml [43]. Similarly a prospective observational study demonstrated that BDG was more accurate than the Candida score and the Colonization Index for early prediction of invasive Candida infection in patients at risk for Candida sepsis [44], with 250 pg/ml identified as the best cutoff value [44]. A decrease in BDG levels during antifungal treatment for invasive candidiasis correlated with a successful clinical response with a positive predictive value of 90 % [45]. However, despite its promising performance for excluding invasive infection and orienting the duration of therapy, the cost of the BDG test and a lack of large validation studies currently limit its use.

Antifungal treatment can be considered as prophylactic (in patients at high risk of fungal infection), empiric (triggered by clinical signs of fungal infection, e.g., persistent fever in the presence of risk factors), preemptive (triggered by microbiological or biomarker evidence of fungus without actual infection), and definitive (after positive microbiological confirmation of strain and sensitivity of fungus).

Prophylactic antifungal agents are recommended in just a few specific situations and most treatments are given on an empiric basis [46]. Although use of preemptive therapy is gaining interest, more studies are needed to better define which patients may benefit from this approach [47] and whether more widespread use of antifungal agents may negatively influence fungal ecology [48]. In a recent study of 241 ICU patients requiring emergency gastrointestinal surgery for intraabdominal infection, preemptive therapy with micafungin was not effective at reducing the development of invasive candidiasis compared with placebo, although there was some suggestion that this may have been the result of the preemptive therapy being administered too late [49].

There are three main groups of antifungals: the azoles, the polyenes, and the echinocandins. The selection of an antifungal regimen is based on multiple factors, including epidemiological data, patient characteristics, hospital setting, fungal strain, site of infection, and safety profiles of the antifungal agents. Although different antifungals can show comparable efficacy in treating candidemia, their differences in terms of pharmacokinetics (PK) and pharmacodynamics (PD) (including fungistatic versus fungicidal activity, need for dose adjustment in case of hepatic or renal failure and drug–drug interactions), toxicity, and selection pressure remain significant and can affect the clinical outcome of fragile patient populations, such as the critically ill. Specific guidelines are therefore available to help direct optimal antifungal choices [12, 46, 50]. The European Society of Clinical Microbiology and Infectious Diseases (ESCMID) guidelines no longer consider fluconazole the drug of choice for invasive candidiasis, and endorse the use of echinocandins as first-line empiric treatment [46]. The rationale for this statement is that, compared with fluconazole, echinocandins show a broader spectrum of activity, fungicidal activity, an excellent safety profile, and fewer drug–drug interactions. The recent Infectious Diseases Society of America (IDSA) guidelines also recommend echinocandins for initial therapy [12]. All three available echinocandins (caspofungin, micafungin, and anidulafungin) penetrate well into the biofilm formed on vascular devices. Nevertheless, fluconazole remains a well-known and well-tolerated antifungal with a significantly lower cost compared with the echinocandins and may still be of value in clinically stable patients who have had no recent azole exposure or have known fluconazole-sensitive organisms [12].