Toxins A and B are the primary virulence factors of Clostridium difficile. Since 2002, an epidemic of C difficile-associated disease with increased morbidity and mortality has been present in Quebec province, Canada. We characterised the dominant strain of this epidemic to determine whether it produces higher amounts of toxins A and B than those produced by non-epidemic strains. We obtained isolates from 124 patients from Centre Hospitalier Universitaire de Sherbrooke in Quebec. Additional isolates from the USA, Canada, and the UK were included to increase the genetic diversity of the toxinotypes tested. Isolate characterisation included toxinotyping, pulsed-field gel electrophoresis (PFGE), PCR ribotyping, detection of a binary toxin gene, and detection of deletions in a putative negative regulator for toxins A and B (tcdC). By use of an enzyme-linked immunoassay, we measured the in-vitro production of toxins A and B by epidemic strain and non-dominant strain isolates. The epidemic strain was characterised as toxinotype III, North American PFGE type 1, and PCR-ribotype 027 (NAP1/027). This strain carried the binary toxin gene cdtB and an 18-bp deletion in tcdC. We isolated this strain from 72 patients with C difficile-associated disease (58 [67%] of 86 with health-care-associated disease; 14 [37%] of 38 with community-acquired disease). Peak median (IQR) toxin A and toxin B concentrations produced in vitro by NAP1/027 were 16 and 23 times higher, respectively, than those measured in isolates representing 12 different PFGE types, known as toxinotype 0 (toxin A, median 848 microg/L [IQR 504-1022] vs 54 microg/L [23-203]; toxin B, 180 microg/L [137-210] vs 8 microg/L [5-25]; p<0.0001 for both toxins). The severity of C difficile-associated disease caused by NAP1/027 could result from hyperproduction of toxins A and B. Dissemination of this strain in North America and Europe could lead to important changes in the epidemiology of C difficile-associated disease.