The performance of the Fungitell assay was investigated in 100 patients with haematological malignancy undergoing chemotherapy who developed antibiotic-unresponsive neutropenic fever (AUNF). Serum beta-D-glucan (BG) concentrations were significantly elevated on the first day of AUNF and all subsequent alternate days to day 10 in 38 patients who developed an invasive fungal infection (IFI) compared to 42 patients remaining free of such infections. The mean and median values of BG were 171.9+/-29.6 and 95.8 pg ml(-1), respectively, for patients with IFI and 64.4+/-17.1 and 32.9 pg ml(-1) for patients with only AUNF (P<0.0001). The differences remained significant over the 10 days despite antifungal therapy. The occurrence of > or =2 sequential concentrations of > or =80 pg ml(-1) ('positive' test) was found to give the best overall option for diagnosis, with an accuracy of 81.3%, sensitivity of 86.8%, positive predictive value of 76.7% and negative predictive value of 86.5%. Of the patients with an IFI, 78% developed a positive test at or before the clinical diagnosis was made -- this occurred at a mean (range) of 1.25 (-14 to +14) days prior to the IFI diagnosis. By starting sampling of blood from the first day of neutropenia rather than from the first day of AUNF, 50% of the patients with subsequent IFI would have been identified 5 days earlier. Increasing sampling to daily from alternate-day frequency did not further improve this earlier timing of an IFI diagnosis. A greater proportion of patients with persistent high levels of BG without overt IFI had severe enterocyte damage or mucositis than those with lower levels of BG without IFI (P=0.002). If the results of the initial BG test had been acted on to change antifungal therapy, discontinuation would have been inappropriate in 30% of patients and would have delayed definitive antifungal therapy. Although the findings for the cohort of patients studied are very useful, there is inter-patient variability in the test's performance. An holistic diagnostic approach is therefore necessary to interpret the test results optimally. Future studies should address this in further detail as well as the impact of empirical antifungal drug use and patient outcome.