The diagnostic performance and usefulness of the Platelia antigen and antibody test (Bio-Rad) was investigated in a prospective study of haematological patients at risk for invasive Candida infections. Among 100 patients, 86 were eligible, of whom invasive candidiasis (IC) occurred in 12 (14%), according to the criteria of the European Organization for Research and Treatment of Cancer/Mycoses Study Group. These included candidaemia due to Candida albicans (one patient) or Candida tropicalis (four patients), and hepatosplenic candidiasis (seven patients). The comparator group of 74 patients included 50 with febrile neutropenia alone and 24 with mould infections. A strategy was developed to determine diagnostic cut-offs from receiver operating characteristic curves with maximal sensitivity and, given this sensitivity, maximal specificity, both being greater than 0. In this patient population, these values were 0.25 ng ml(-1) for mannan (M) and 2.6 arbitrary units ml(-1) for anti-mannan (AM), which are lower than those recommended by the manufacturer. All patients developed at least one positive diagnostic M or AM result during the 10 days of persistent febrile neutropenia (PFN). The optimal overall performance was found when two consecutive positive tests for both M and AM were used [sensitivity, specificity, positive predictive value and negative predictive value (NPV) (95 % confidence intervals) of 0.73 (0.39-0.94), 0.80 (0.69-0.89), 0.36 (0.17-0.59) and 0.95 (0.86-0.99), respectively]. There was a positive correlation of M with beta-D-glucan (r=0.28, P=0.01). The first positive M test was found up to a mean+/-sd of 8.8+/-8.5 (range 2-23) days prior to a clinical/mycological diagnosis of IC. Day-to-day variation in quantitative M levels was high. High-level AM responses were delayed until leucopenia resolved. The low specificities of the test performance may have been due to some of the comparator patients having subclinical Candida infections as evidenced by the high incidence of colonization among them (60% had a colonization index of >or=0.5). The high NPVs suggest that the tests may be particularly useful in excluding IC. It is feasible to explore the use of serial measurements of M and AM as part of a broader diagnostic strategy for selecting PFN patients to receive antifungal drug therapy.