The intra- and interlaboratory reproducibilities of a commercial sandwich enzyme-linked immunosorbent assay (ELISA) for the detection of Aspergillus galactomannan in serum (Platelia Aspergillus; Sanofi Diagnostics Pasteur, Marnes-La-Coquette, France) were evaluated in six laboratories of university hospitals. Twenty serum samples were obtained from 12 neutropenic patients including 6 with invasive aspergillosis. These samples were blinded and sent to each center together with eight blinded ELISA-negative serum samples spiked with known concentrations of galactomannan. The centers were provided with ELISA microtiter plates from a single batch and a detailed protocol. Ten clinical samples showed ELISA reactivity, while 10 samples were ELISA negative. The mean coefficient of variation (CV) of the optical density values was 4.24% within a single assay and 25.6% between runs. The interassay CV of the ratios for the serum samples tested was 18.6%. Analysis of ordinal interpretation of the ELISA result (i.e., negative, gray zone, or positive) showed excellent reproducibility. Recalculation of the cutoff values for positive and negative samples suggested that the cutoff level recommended by the manufacturer could be lowered from 1.0 to 0.8 for negative samples and from 1.5 to 1.0 for positive samples. The intra- and interlaboratory reproducibilities were excellent when the ELISA results were interpreted as ordinal data, but considerable variation in optical density values and, to a lesser extent, in the ratios for the serum samples tested, was observed between runs. High assay variability was also found for serum samples spiked with known concentrations of galactomannan. Therefore, antigen titers in serum samples from a single patient, measured in different runs, should be compared with caution.