Invasive candidiasis (IC), occur in 10 to 15% of ICU patients and are associated with an overall mortality of 30 to 50% 1,2 . In this context, it is especially candidemia or intra-abdominal candidiasis (CIA). One of the factors widely recognized as capable of improving their prognosis on early antifungal treatment adapted 3.4 . However, diagnostic means for confirming a CI always based on the sterile sites of cultures. These cultures are very sensitive (at best 50% in candidemia) and too late (between 1 and 3 day period for the detection of yeasts) 5 .
To reduce mortality, a significant number of ICU patients with risk factors (heavy digestive surgery, colonization multisite, mechanical ventilation, central venous catheter, parenteral nutrition, acute renal failure, prolonged antibiotic therapy) receive empiric antifungal therapy when they present a picture of severe sepsis possibly related to a deep candidiasis. Some studies have shown that early treatment strategy, preceding the mycological literature, improves survival 6 while others have not found benefit 7.8 . As a result, nearly 70% of patients receiving antifungal therapy in ICU have no documented infection 9 . Besides, efficiency discussed these empirical treatments pose selection pressure problems and toxicity 10 . Faced with the rapid emergence of resistance, particularly those of Candida glabrata to echinocandins and azoles, the question of rationalizing the use of antifungal becomes crucial.
To improve the selection of patients who may benefit from early antifungal treatment, scores or predictive models were developed and evaluated. They variously associated risk factors and clinical scores of colonization index toCandida spp. 11-16 . These scores have a high sensitivity (77% -100%) but lack specificity (48% -74%) and especially have an interest in their good negative predictive value (72% -100%) leading to selection of patients who not require antifungal therapy. They also allow to argue stopping antifungal treatment would have been instituted unnecessarily 14.16 to 18 . In practice, this approach based on negative and leading to drug therapy, concerns only very few patients in the ICU 18 .
The detection of certain biomarkers (BM) of Candida infection has been developed to overcome the drawbacks of diagnostic methods based only on culture. Ideally, the BM should allow earlier diagnosis but also have prognostic value while evaluating the response to treatment. These are either BM circulating antibody response to antigens of cell wall of yeast, or of cell wall polysaccharides such as mannan or BD-glucans. Molecular methods, mainly PCR, to detect DNA of Candida spp. in the blood or in the serum. Promising methods using nuclear magnetic resonance (T2MR), mass spectrometry associated with PCR or spectroscopy are in development 19 .
The test for the joint detection of anti-mannan antibodies and mannan was recently amended (Mannan / AntiMannan Platelia CandidaAg Plus, BioRad). The evaluation of new kit for ICU patients shows that despite a lowering of mannans positivity threshold to 50 pg / ml, the positive and negative predictive values were 55% and 96% respectively. The ability of mannan antigens to distinguish colonization multisite and superficial candidiasis candidiasis remains unclear. The performance of anti-mannan antibodies alone are insufficient and their association with manannes antigens has no interest in this study 20 .
The CATGA antibody ( Candida albicans germ tube antibody) are directed against cell wall proteins of Candidaspp. This test has been little evaluated in ICU patients but showed the capacity to differentiate colonization from infection. It is not changed by antifungal treatment and may also have prognostic value 21 .
The detection of DNA of Candida spp. has generated much interest and overall performance demonstrate a sensitivity of 80 to 100% and a specificity of 90 to 100% 19 . In addition to allowing the identification of the species, its main interest is the detection of deep candidiasis with candidemia not like the CIA with a sensitivity of 88% against 17% for blood culture 22.23 . Standardization efforts and prospective validation of PCR tests are needed before use in clinical practice.
BD-glucan (BDG) are polysaccharide parietal markers and pan fungal released in serum by Aspergillus, Candida, Pneumocystis, Fusarium ... In the context of the diagnosis of IC, they were evaluated especially in critically ill patients after gastrointestinal surgery with high risk of CI and particularly CIA. The sensitivity and the specificity varies from 62% respectively 87% and 73% to 98% 20,24 . In patients at very high risk of CIA (recurrent necrotizing intestinal perforation or peritonitis) allow the BDG diagnosed 5 days before the culture, are higher in severe patients and continue to increase in case of treatment failure. They also distinguish colonization and infection 25 . Always in patients after gastrointestinal surgery, a high rate of BDG (> 259 pg / ml) associated with positive CATGA allows discriminating colonization and infection 26 . In more heterogeneous populations at risk for candidemia more than CIA, the only BDG lack specificity with respective positive and negative predictive values of 41% to 72% and 99%. Similarly, in moderate risk patients (incidence 10-15%), preemptive treatment with caspofungin established after two consecutive positive BDG brings no benefit 27 . In non-surgical patients, the association with mannan antigens or colonization scores did not improve the performance of BDG 20.28 . The disadvantages of BDG test that limit its routine use are still uncertainties about the conditions for sampling and performing the test and its cost.
In conclusion, the BM performance available: mannan antigens and anti-mannan antibodies or CATGA, PCR or BDG vary depending on clinical situations and CHD risk levels. The BDG (alone or combined with CATGA) and PCR can anticipate the initiation of antifungal therapy in patients after gastrointestinal surgery at very high risk (incidence> 30%). BM These appear even more powerful and useful than the CIA risk is high, the situation in which crops are less sensitive. In this situation and in addition to their diagnostic value, the BDG have a prognostic value and to assess response to treatment. BM performance in the selection of patients to treat early decline in populations of non-surgical ICU patients at lower risk of CI. However, in these situations, their good negative predictive value avoids unnecessary treatment. It is important to stress that these BM may only be used in addition to the methods of cultures which remain essential for species identification and detection of resistance of Candida spp.