Amphotericin B has been the standard therapy for invasive aspergillosis since its introduction in 1957. It is only moderately effective. Many susceptibility tests have been used but little variation has been noted between strains. We have studied three strains of Aspergillus fumigatusand one of Aspergillus terreusin a neutropenic mouse model of invasive aspergillosis and attempted to correlate the variable efficacy in vivowith MICs generated by over 30 different susceptibility test formats. One strain of A. fumigatus(AF65) and the strain of A. terreus(AT49) were 'resistant' and the remaining two strains of A. fumigatus(AF210 and AF294) were 'susceptible' in vivo. Only AT49 had elevated MICs of amphotericin (MIC 2 mg/L) by 41 of 54 in vitrotesting systems. With each test format, including Etest, there was no distinction between MICs obtained for AF65, AF210 and AF294 (MICs 0.125-64 mg/L depending on the test). Thus despite extensive efforts we have been unable to correlate susceptible test results with in vivooutcome in A. fumigatusbut we have with A. terreus, with some test formats. This suggests that, at present, amphotericin B susceptibility testing of A. fumigatus is of limited clinical value and further work needs to be done to find testing systems that can identify the 'resistance' documented in vivo.