Introduction

Invasive candidiasis (IC) is a major cause of morbidity and mortality in critically ill patients [1]. Critically ill patients have several risk factors for IC development such as broad-spectrum antibiotic treatment, disruption of natural barriers due to surgery or invasive devices, impaired immunological function, and fungal colonization [2, 3]. IC comprises both candidemia, which is commonly considered the most frequent type in critically ill patients, and deep-seated candidiasis (e.g., intra-abdominal candidiasis) that is also relatively common in specific subgroups of patients (e.g., post-surgical abdominal patients) with or without concomitant candidemia [4]. Candida spp. are the most common fungal microorganisms encountered in the Intensive Care Unit (ICU), ranking the third isolated pathogens in ICU patients and the fourth in nosocomial bloodstream infections (BSI) [5, 6]. In the EUROBACT international cohort study, fungal bloodstream infections (BSI) accounted for 7.8% of monomicrobial infections [7]. Attributable mortality due to Candida spp. infections is high, ranging from 42 to about 60% according to different studies and depending of the development of septic shock in studied populations [5, 8, 9]. Candidemia significantly increases also in-hospital length of stay and costs, with a mean increase in hospital charges of nearly $40.000 per case [10]. Several observational studies described a strict correlation between time of antifungal treatment initiation and patients’ mortality [8, 11,12,13,14]. However, other studies did not confirm these findings with severity of illness being one of the strongest predictors of mortality in patients with proven IC [15,16,17,18]. Blood cultures are the gold standard for IC diagnosis but the microbiological identification of fungal pathogens and their antifungal susceptibility is time-consuming and usually occurs late in the clinical course of patients with suspected IC [19, 20]. Indeed, the sensitivity of blood cultures for the identification of Candida is not high: almost 50% of patients may have negative results [21, 22]. Antifungal drugs and source control are cornerstones of treatment of IC [1, 23]. In light of these considerations, it is common for clinicians to administer antifungal agents before definitive diagnosis of IC [17, 24, 25]. Widespread use of antifungal treatment is not without drawbacks since antifungal drugs are costly and there may be a link with the development of resistance to antifungals [26,27,28]. The aim of this review is to provide an updated summary of evidence from trials, systematic reviews, and guidelines about strategies to prevent or treat appropriately IC in non-neutropenic critically ill patients before definitive microbiological diagnosis along with critical insights for its interpretation.

Untargeted Antifungal Treatment Strategies

Aims and Definitions

From more than 30 years, clinicians and researchers have tried to prevent/treat early IC in critically ill patients with different therapeutic approaches. Three main antifungal strategies have been described in literature and used in clinical practice with this aim (Fig. 1) [3, 29, 30]: (1) Antifungal prophylaxis has been defined as the administration of antifungal drugs to patients without signs or symptoms of IC but with risk factors for its development (e.g., central venous catheter, parenteral nutrition, dialysis, broad-spectrum antibiotics, fungal colonization). This strategy has been frequently applied in specific subgroups of patients at risk of IC development such as post-surgical abdominal patients or with the help of prediction scores (e.g. Candida score, Ostrosky–Zeichner score, Paphitou score, FIRE score) to optimize costs and benefits (see Table 1) [31, 32]; (2) Empiric treatment has been defined as the administration of antifungal drugs to patients presenting signs and symptoms of infection potentially due to fungi and at risk of IC development (e.g., febrile ICU patients despite broad-spectrum antibiotics, septic patients with potential intra-abdominal focus of infection); (3) Pre-emptive antifungal treatment is defined as a treatment triggered by evidence of fungal infection, basing on “surrogate marker” or non-culture diagnostic tests, without definitive microbiological identification of fungal pathogen (e.g., positive biomarkers 1-3 beta-D-glucan, mannan-antimannan antibodies, polymerase chain reaction assays) [21, 33,34,35,36,37]. This relatively new strategy aims to narrow the large target population of prophylaxis and to reduce the time of initiation of empiric treatment. Notably, the terminology for antifungal strategies is not standardized in literature leading to difficulty in classifying the interventions in both clinical practice and research. The common point of these strategies is that they are “untargeted” since they are applied without definitive microbiological IC diagnosis and pathogen identification. The number of patients treated with antifungal drugs without definitive diagnosis of IC is high. In a 1-day cross-sectional cohort study performed in 169 ICUs in France and Belgium, systemic antifungal therapy was used in 7.5% of currently admitted patients [24]. Two thirds of these patients had no documented invasive fungal infection (IFI). In a 1-year prospective observational multicenter cohort study performed in 87 ICUs in France, 835 non-neutropenic patients received systemic antifungal treatment for proven or suspected IC. Among the 544 receiving antifungals for a suspected infection, 423 did not have documented infections at 28 days after inclusion [17].

Fig. 1
figure 1

The picture shows untargeted antifungal treatment strategies and their increasing specificity

Full size image
Table 1 Predictive scores for Candida infections
Full size table

Evidence From Randomized Controlled Trials

Untargeted antifungal treatment strategies have been studied for nearly three decades, with the first randomized controlled trial (RCT) reported in major literature dating from 1987 [38]. Early RCTs studied the effect of azoles, mainly fluconazole, in both medical and surgical critically ill patients. In 2006, a Cochrane systematic review summarized available evidence from RCTs on this topic [39]. At that time, untargeted antifungal treatment strategies were not well defined and the intervention under investigation in that review was generically defined as “prophylaxis.” The review included 12 RCTs, 8 with fluconazole and 4 with ketoconazole, compared to placebo or no intervention, for a total of 1606 non-neutropenic patients. Antifungal treatment reduced total mortality by about 25% (relative risk—RR −0.76, 95% CI 0.59 to 0.97) and IFI rate by about 50% (RR 0.46, 95% CI 0.31 to 0.68). Interestingly, in the subgroup analysis including trials with more than 75% of post-surgical patients, the effect of the intervention was still significant. These results had an important impact in clinical practice due to the high magnitude of the effect in terms of survival and reduction of IFI development. However, most of the included studies did not have multicenter design, and a grading of the quality of evidence was not performed. In the following decade, subsequent large multicenter RCTs provided different results studying, in most cases, the effect of newer drugs. Schuster et al. randomized 270 adult ICU non-neutropenic patients with fever despite broad-spectrum antibiotic treatment to receive 800 mg fluconazole daily or placebo [40]. All patients had a central line and an Acute Physiology and Chronic Health Evaluation (APACHE) II score more greater than 16. The study was performed in 25 ICUs in the USA. The composite primary outcome, consisting on resolution of fever, absence of IFI, no discontinuation because of toxicity, and no need for a non-study systemic antifungal medication, was not significantly different between intervention and control group (RR 0.95–95% CI 0.69 to 1.32). Moreover, antifungal treatment did not reduce mortality or incidence of IC. Ostrosky–Zeichner et al. included 222 ICU patients in a multicenter RCT (15 ICUs in the USA) comparing caspofungin to placebo [41]. All included patients had an ICU stay of at least 3 days, were mechanically ventilated, received antibiotics, had central venous lines, and one another additional risk factor among common ones. The primary outcome was the incidence of proven or probable IC in patients who did not have the disease at baseline. Neither the primary outcome nor mortality was significantly different between the intervention and the control group (16 vs 9.8%; P = .14). Knitsch et al. randomized 252 patients with localized or generalized intra-abdominal infection requiring emergency surgery and ICU stay [42]. Patients received micafungin or placebo. The primary outcome was the incidence of IC. There was no significant difference in the primary outcome among intervention and control group (11.1 vs 8.9%—difference 2.24%; 95% CI −5.52 to 10.20) as well as for mortality. In 2016, Cortegiani et al. published a Cochrane systematic review summarizing available evidence from RCTs about the concept of “untargeted” antifungal treatment in non-neutropenic patients [38, 43]. The primary outcomes were mortality and incidence of proven IFI. Differently from the one published in 2006, this review provided an evaluation of quality of the evidence according to the GRADE approach. Moreover, authors modified the definition of one of the primary outcome (proven IFI) excluding the isolation of fungi from the respiratory tract of included patients and considering it as colonization, in line with current evidence [29]. This last Cochrane review included 22 studies for a total of 2761 patients. All the included studies evaluated IFI from Candida spp. There was moderate-grade evidence that untargeted antifungal treatment did not significantly reduce or increase total all-cause mortality (RR 0.93, 95% CI 0.79 to 1.09) compared to placebo or no antifungals. On the other hand, untargeted antifungal treatment significantly reduced the risk of developing proven IFI (RR 0.57, 95% CI 0.39 to 0.83) but the quality of the evidence was low. In the subgroup analysis evaluating trials with more than 75% of post-surgical patients, the results were consistent with no mortality benefit but reduced risk of proven IFI. Interestingly, in the subgroup analysis for type of intervention, antifungal prophylaxis was associated with no significant effect on mortality but with a reduced risk of proven IFI, whereas empiric treatment did not reduce either mortality or proven IFI [44, 45]. Notably, only one trial, with questionable design, evaluated empiric treatment based on beta-D-glucan. The authors also performed a subgroup analysis according to drug class: azoles were associated with reduced mortality and proven IFI differently from echinocandins. However, more studies evaluated the effects of azoles (12) than echinocandins (4). At the end of 2016, Timsit et al. published the results of a multicenter randomized double-blind placebo-controlled trial (EMPIRICUS) investigating the effect of empiric antifungal treatment in non-neutropenic critically ill patients with ICU-acquired sepsis, multiple-site Candida colonization, multiple organ failure, and receiving broad-spectrum antibiotics [46]. The study was performed in 19 ICUs in France. Patients were randomized to receive micafungin or placebo. The primary outcome was survival without proven IFI at 28 days from randomization. Two hundred eighty patients were randomized, with a high grade of critical illness (mean SAPS II score of 48 and median SOFA score of 8). Regarding the primary outcome, no significant difference was found between groups (68 vs 60.2%—hazard ratio 1.35, 95% CI 0.87–2.08). Moreover, no difference was found in survival at 28 and 90 days whereas the number of patients developing new IFI was significantly higher in the placebo group compared to micafungin group.

Clinical Practice Guidelines

Two international scientific societies (European Society of Clinical Microbiology—ESCMID—and the Infectious Disease Society of America—IDSA) published clinical practice guidelines concerning management of Candida diseases in non-neutropenic patients basing on available literature. In the latest ESCMID guidelines, published in 2012, no specific recommendation according to the GRADE approach for quality assessment of body of evidence was provided [29]. Antifungal prophylaxis with fluconazole was recommended in patients who underwent abdominal surgery and had gastrointestinal perforations or anastomotic leakages. Concerning empiric treatment, no recommendation was given about drugs or optimal timing. However, it was stated that early treatment is presumably linked to higher survival rate. Echinocandins were preferred over fluconazole, which was somehow downgraded. Pre-emptive treatment basing on beta-D-glucan was marginally supported. The IDSA clinical practice guidelines published in 2016 provided an assessment of quality of evidence and strength of recommendation with the GRADE methodology [23, 47]. Regarding antifungal prophylaxis, guidelines stated that fluconazole 800 mg as loading dose and then 400 mg daily (with an echinocandin as alternative) could be used in high-risk patients in adult ICUs with a high rate (> 5%) of IC (week recommendation; low-quality evidence). Empiric treatment was supported by a strong recommendation and moderate-quality evidence and should be considered in critically ill patients with risk factors for IC and no other known cause of fever. The decision to start empiric therapy should be based on clinical assessment of risk factors, surrogate markers and/or culture data from nonsterile site. If present, risk factors for IC in association to septic shock should trigger empiric treatment initiation as soon as possible. An echinocandin should be the first-line drug (strong recommendation; moderate-quality evidence) with fluconazole being an acceptable alternative for hemodynamically stable patients without recent azole exposure or colonization with azole-resistant Candida species (e.g., Candida krusei, Candida glabrata). The suggested duration of empiric treatment is 2 weeks, and in patients without clinical response after 4–5 days and without subsequent evidence of IC, treatment could be stopped—a practice known as de-escalation—(strong recommendation; low-quality evidence) [48]. In suspected intra-abdominal candidiasis, empiric treatment, in association with source control, should be considered basing on clinical signs and risk factors, including recent abdominal surgery, anastomotic leaks, or necrotizing pancreatitis (strong recommendation; moderate-quality evidence). It should be noted that IDSA 2016 guideline did not take into account evidence from the study by Knitsch et al., 2016 Cochrane systematic review and EMPIRICUS trial due to timing of preliminary literature search. The use of surrogate marker to guide antifungal treatment in ICU was also taken into account with 1-3-beta-D-glucan being the most studied but not in randomized trials. Beta-D-glucan is a cell wall component of several fungi such as Candida spp., Aspergillus spp., and Pneumocystis jiroveci. For this reason, a positive test is not specific for IC. Although, some data showed that it can anticipate the results of blood culture shortening the time to antifungal treatment initiation and its levels may correlate with response to antifungals, especially with serial measurements, the major limitation of 1-3-beta-D-glucan is the low positive predictive value in ICU patients [23, 48, 49]. However, the negative predictive value seems to be adequate to help rule out the diagnosis of IFI [50, 51]. It should also be noted that RCTs evaluating its performance are lacking.

Potential Interpretations and Insights

In light of recent data, some clinical questions on untargeted antifungal strategies may rise, especially on the possible implementation of treatment strategies together with biomarkers to improve diagnostic specificity and stop antifungal treatments. The main clinical questions may then be: how is it possible to have a lack of benefit in terms of survival along with a reduction in the incidence of IC? What are the reasons for these findings? Should we continue to administer antifungal drugs to our critically ill non-neutropenic patients before definitive diagnosis of IC? Since IC is associated with high morbidity and mortality, the positive effect of untargeted antifungal strategies in terms of risk of proven IC along with a non-significant effect on mortality may be seen as a paradox. The interpretation of these findings is complex and may be only speculative [52,53,54,55,56]. The widespread use of antifungals is linked to the development of resistance [57]. Echinocandin resistance is an increasing microbiological phenomenon and, as a consequence of increased selective pressure, an epidemiological change towards less susceptible Candida species (e.g. C. glabrata) may be seen [27]. Moreover, acquired resistance is an emerging significant problem, with the spread of genetic mutation in “hot-spot” regions especially in C. glabrata (e.g. FKS1, FKS2) [28, 58]. Multicenter trials published in the last decade significantly increased the number of included patients and improved the generalizability of results and the quality of body of evidence on this topic changing the overall effect on mortality. The research on this topic covers a wide period of time, nearly three decades, during which the care of critically ill patients have significantly improved, especially for the management of infections and sepsis. The consequence of this process may have been a blunted effect of the reduction of IC on mortality [52].

IFI are a frequent complication of immunosuppression in transplantation recipients, patients with hematologic malignancies, and neutropenia. The incidence trend of IFI has substantially increased over time, even in ICU patients not belonging to these categories, due to improved care of both medical and surgical critical conditions [59]. Although antifungal drugs are very effective in killing fungi in body sites leading to a reduced proven IFI, many non-neutropenic patients may die despite of effective treatment [60]. Nowadays, there is an established role of impaired immunological function in septic patients, even without “classic” mechanisms of immunosuppression [61]. In animal models, impaired immunity due to T cell exhaustion increases fungal sepsis-related mortality. Recently, Spec et al. published an observational study evaluating the immunophenotype of non-neutropenic patients with candidemia, compared to non-septic critically ill controls [62]. Both CD4 and CD8 T cells presented an immunophenotype consistent with immunosuppression (markers of cell exhaustion—PD1 and PDL1—and downregulation of positive co-stimulatory molecules). It may be argued that the reduced immunological function may have a causative role on the lack of benefit of highly effective antifungals or it may rather represent a marker of severe underlying disease [63]. Immunoadjuvant treatments to reverse T cell exhaustion and boost immunity may be therapeutic goals for the near future.

Although the quality of evidence from RCTs on this topic has substantially improved during the last decade, it is not high [38]. It is unlikely that the quality of body of evidence will further improve due to the fact that these antifungal interventions have been extensively studied since many years; several RCTs globally included more than 2500 critically ill patients and the costs to perform other RCTs would be hardly justified. However, since “classic” untargeted antifungal strategies have failed to improve mortality, future research should evaluate strategies combining risk factors, available biomarkers, and risk scores to optimize patients’ selection, timing of treatment and cost/benefit balance [64, 65]. Several non-culture-based assays and surrogate markers are available nowadays to select target patients’ population, to reduce turn-around time to IC diagnosis, and to improve the diagnostic performance of classic culture-based methods [22, 65]. A full description of their characteristics is beyond the aim of this review and can be retrieved elsewhere [23, 65]. Candida albicans germ tube antibody (CAGTA), mannan-antigen and antimannan antibodies, polymerase chain reactions, and matrix-assisted laser desorption ionization-time of flight mass spectrophotometry (MALDI-TOF MS) have been studied alone or in combination with promising results but further data are needed to clarify their diagnostic performance, clinical usefulness and cost/benefit balance [65]. New promising tests are those using T2 magnetic resonance (T2MR) for the rapid detection of Candida species from whole-blood samples. T2MR can rapidly detect the presence of molecular (e.g., DNA, proteins) in a sample without the need for extraction and purification of target molecules from the sample [66]. Recently, a clinical trial compared T2MR to blood culture for the diagnosis of candidemia in 1801 patients. T2MR demonstrated an overall sensitivity of 91.1%, specificity of 99.4%, an estimated negative predictive value of about 99%, and a mean time of 4.4 h for detection and species identification [67]. Further studies should evaluate its large-scale applicability within diagnostic-therapeutic algorithms. Antifungal stewardship is a complex of interventions ranging from patients’ selection to early institution of effective antifungal treatment, implementation of methods for early diagnosis of IFI, outcome monitoring, promotion of audits, and educational interventions [68, 69]. The goals of these programs are better use of resources, reduce resistance to antimicrobials, and optimize patients’ outcome. The reported experience with the implementation of antifungal stewardship is scarce compared to that of antibacterial initiatives. The obstacles against a wider adoption of antifungal stewardship are mainly due to the current difficulty in proper patient selection [70]. Indeed, given the described issues on early identification of IC, clinicians should balance the risks of delayed antifungal administration against the futile exposure to antifungals, with associated increased risk of resistance, costs, and adverse events.

Conclusions

After more than three decades of research and clinical application, a widespread use of untargeted antifungal treatment strategies seems to be no longer justified and diagnostic specificity is the key of success. Clinicians should continue to start systemic antifungal treatment before definitive diagnosis of IC taking into account all available sources of information, later aiming at de-escalation: patients’ characteristics, risk factor assessment (including fungal colonization) with eventual calculation of clinical scores, and use of surrogate biomarkers, if available (Fig. 1) [48, 53, 65]. The goals of this comprehensive strategy should be to optimize patients’ selection, reduce exposure to antifungals, limit development of resistance to antifungals, and reduce costs. Further research is needed to strengthen the evidence on new combined diagnostic-therapeutic antifungal strategies in accordance with antifungal stewardship principles and de-escalation.