The recent advent of an improved commercial serum enzyme-linked immunosorbent assay (ELISA) for the detection of circulating galactomannan (GM), a major constituent of Aspergillus cell walls, has contributed to the diagnosis of invasive aspergillosis (IA) in many haematology and transplant centres. However, the optimal threshold for positivity remains a matter of debate. We prospectively evaluated the impact of lowering the cut-off in 124 neutropenic episodes with a high pretest probability for IA. Two new cut-off points, lower than previously accepted, are proposed: (a) a 'static' cut-off at 0.8 and (b) a 'dynamic' cut-off at 0.5. A single assay with an optical density (OD) index > or = 0.8 warrants the initiation of anti-Aspergillus therapy. A further lowering of the 'static' threshold seems not clinically feasible given the drop in positive predictive value (PPV). However, the demonstration of at least two sequential sera with an OD > or = 0.5 ('dynamic' threshold) increased the specificity and the PPV to 98.6% and the efficiency to 98%. Applying both cut-offs to a subgroup of 21 'possible' fungal infections further identified and upgraded six cases of IA. However, the clinical benefit of lower cut-offs (particularly for earlier diagnosis) depends upon the kinetics of antigenaemia and the intensity of serum sampling.