We recently attempted to clarify an increased incidence of Clostridium difficile-associated diarrhoea (CDAD) in our hospital by arbitrarily primed polymerase chain reaction (AP-PCR) typing of isolates from 147 consecutive patients collected during a 12 month period (Wullt et al. J Hosp Infect 1999;43:265-273). In the present study we compared the results based on previous AP-PCR data with those based on recent PCR ribotyping of the same isolates and re-analysis of a subset of isolates by AP-PCR typing. The pattern of PCR ribotypes was similar among inpatients and outpatients. A cluster of three closely related PCR ribotypes, related to those of the serogroup H and A8 type strains, dominated and comprised 31% of inpatient and 28% of outpatient C. difficile isolates. The apparent nosocomial transmission rate among inpatients with CDAD was only 9% by AP-PCR typing compared with 18 or 36% by PCR ribotyping depending on the definition used (proportion of patients sharing C. difficile type and ward within two or 12 months). Corresponding rates for all CDAD patients were 5% by AP-PCR and 11 or 21% by PCR ribotyping. Thus, most CDAD patients apparently became ill due to their endogenous strain of C. difficile. Because of the low concordance between the two typing methods the proportion of patients fulfilling the criteria for nosocomial transmission by both methods was only 1%. Re-examination of isolates from patients with recurrences revealed a reproducibility problem with AP-PCR typing. We conclude, that of these two PCR-based options for typing of C. difficile PCR ribotyping offers a superior experimental robustness compared with AP-PCR typing.