Non-culture-based biomarkers may improve diagnosis and antifungal treatment (AFT) of invasive candidiasis (IC). We evaluated an antifungal stewardship programme (AFSP) in a prospective intensive care unit (ICU) study, which included T2Candida and Candida mannan antigen (MAg) screening of patients with sepsis and a high risk of IC. Patients with non-neutropenic sepsis and a high risk of IC from two large tertiary ICUs were prospectively included, during a one-year period. IC was classified as proven, likely, possible or unlikely. The AFSP, diagnostic values of T2Candida and MAg, and the consumption of antifungals were evaluated. An amount of 219 patients with 504 T2Candida/MAg samples were included. IC was classified as proven in 29 (13.2%), likely in 7 (3.2%) and possible in 10 (5.5%) patients. Sensitivity/specificity/PPV/NPV values, comparing proven/likely versus unlikely IC, were 47%/100%/94%/90% for BC alone, 50%/97%/75%/90% for T2Candida alone, and 39%/96%/67%/88% for MAg alone. For the combination of T2Candida/MAg taken ≤3 days after AFT initiation, sensitivity/specificity/PPV/NPV was 70%/90%/63%/93%. T2Candida/MAg contributed to early (<3 days) AFT initiation in 13%, early AFT discontinuation in 25% and abstaining from AFT in 24% of patients. No reduction in overall use of AFT during the study period compared with the previous year was observed. An AFSP based on T2Candida and MAg screening contributed to a reduction of unnecessary treatment, but not overall AFT use. The diagnostic performance of T2Candida was lower than previously reported, but increased if T2Candida was combined with MAg.