Invasive candidiasis is associated with high mortality rates (35% to 60%), similar to the range reported for septic shock. The most common types include candidemia, frequently observed in immunocompromised patients, and noncandidemic systemic candidiasis, which constitutes the majority of cases in critically ill patients. However, they are difficult to prove and a definite diagnosis usually occurs late in the course of the disease, thus contributing to their bad prognosis. Early empirical treatment improves the prognosis and currently relies on the positive predictive value (PPV) of risk-assessment strategies (colonization index, Candida score, predictive rules) based on combinations of risk factors, but it may have also largely contributed to the overuse of antifungal agents in critically ill patients. In this context, non- culture-based diagnostic methods, including specific and nonspecific biomarkers, may significantly improve the diagnosis of invasive candidiasis. Candida DNA and mannan antigen/antimannan antibodies are of limited interest for the diagnosis of invasive candidiasis as they fail to identify noncandidemic systemic candidiasis, despite early positivity in candidemic patients. The utility of 1,3-beta-D-glucan (b-D-glucan), a panfungal cell wall antigen, has been demonstrated for the diagnosis of fungal infections in immunocompromised patients. Preliminary data suggest that it is also detectable early in critically ill patients developing noncandidemic systemic candidiasis. To take advantage of the high negative predictive value of risk-assessment strategies and the early increase in specific fungal biomarkers in high-risk patients, we propose a practical 2-step approach to improve the selection of patients susceptible to benefit from empirical antifungal treatment.