Persistent or recurrent fevers of unknown origin (PFUO) in neutropenic patients on broad-spectrum antibiotics have traditionally been treated with empirical antifungal therapy (EAFT). The lack of survival benefit seen with the use of amphotericin B deoxycholate (AmB-D) as EAFT has been attributed to its toxicities. More recently, newer, less toxic and more expensive antifungal agents such as the lipid formulations of AmB, the newer azoles (fluconazole, itraconazole and voriconazole) and caspofungin have been analysed in a number of EAFT trials. Compared with AmB-D the newer agents have superior safety but are of equivalent efficacy. This lack of survival advantage is related to the fact that the trigger for commencement of EAFT is late and non-specific. Thus, alternative approaches are required. New sensitive serological and molecular tests for the detection of Aspergillus antigens and genomic DNA have been developed and evaluated in accuracy studies. These tests have been incorporated into management strategies (i.e. pre-emptive strategies) to direct antifungal therapy. The pre-emptive approach has been shown to be safe and feasible but its impact on clinically important patient outcomes such as survival is less clear. Other advances include the introduction of effective, non-toxic mould-active antifungal prophylaxis and patient risk-group stratification. In this paper we provide new evidence-based algorithms for the diagnosis and treatment of PFUO in adult patients undergoing stem cell transplantation and chemotherapy for haematological malignancy which incorporate these newer diagnostic tests and are directed by the risk category of the patient and type of antifungal prophylaxis the patient is receiving.