We have evaluated the contribution of the 1,3-beta-d-glucan (BG) assay for the screening of invasive fungal infections (IFIs) in patients with haematological malignancies. Serum samples from patients at risk of IFI were collected twice a week and retrospectively tested using the BG assay. BG screening was performed on 1143 samples from 91 patients during 104 anticancer treatment cycles. Proven and probable cases of IFI occurred in 9 (8.7 %) treatment cycles. Depending on the criterion of positivity used (1x >60 pg ml(-1), 1x >80 pg ml(-1), 2x >60 pg ml(-1) or 2x >80 pg ml(-1)) the sensitivity and specificity were 89, 89, 67 and 44 %, and 20, 48, 33 and 56 %, respectively. Although the test was marked as positive in 82, 68, 54 and 45 % of all the treatment cycles, in the majority of cases, these positivities were probably false. The major limit of the BG test was an extremely low positive predictive value (10 to 12 %). We have analysed mucositis, candida colonization, bacteraemia, use of antimicrobials, erythrocyte and thrombocyte filtered blood products, collecting tubes or sampling via venous catheters. Even though no factor is a major source of BG, it could at least partially influence BG assay performance. Thus, BG detection has a limited usefulness as a screening method for IFIs in patients with haematological malignancies.