In this prospective multicentre study, an enzyme-linked fluorescent assay (VIDAS CDA2; bioMérieux), an enzyme-linked assay [Premier Toxins A and B (PTAB); Meridian] and an in-house real-time PCR amplifying the tcdB gene were compared with the cell cytotoxicity assay used as the 'gold standard' for diagnosis of Clostridium difficile-associated diarrhoea (CDAD). Faecal samples from patients with a request for C. difficile diagnosis and samples from patients with diarrhoea hospitalized for at least 72 h were collected for 3 consecutive months from four university medical centres in The Netherlands. In total, 547 faecal samples were obtained from 450 patients. Of 540 samples available for all of the assays, 84 (15.6 %) showed a positive result in one or more assays. The cell cytotoxicity assay was positive in 31 samples (5.7 %) from 28 patients. A diagnosis of CDAD was not considered by the physician in 5 (23.8 %) of 21 patients with CDAD who were hospitalized for at least 72 h. Compared with the cell cytotoxicity assay, the sensitivity of VIDAS, PTAB and PCR was 83.9, 96.8 and 87.1 %, respectively. The specificity of VIDAS, PTAB and PCR was 97.1, 94.3 and 96.5 %, respectively. The positive and negative predictive values for VIDAS, PTAB and PCR were 63.4 and 99.0 %, 50.9 and 99.8 %, and 60.0 and 99.2 %, respectively. Of 61 samples that were positive in one, two or three of the assays, 56 were available for discordance analysis. Discordance analysis was performed by culture of toxinogenic strains. The concordance of VIDAS, PTAB and PCR with culture was 53.6 % (30/56), 55.4 % (31/56) and 71.4 % (40/56), respectively. It was concluded that real-time PCR had the highest concordance with toxinogenic culture and is therefore the preferred method for diagnosing CDAD in faecal samples. It was also concluded that diagnosis of patients with diarrhoea who have been hospitalized for more than 72 h should focus mainly on the detection of C. difficile, irrespective of the physician's request.