Objective: To scrutinize published sensitivity estimates obtained using questionable gold standards by comparing sensitivities of culturing Clostridium difficile in commercially available media followed by enzyme immunoassay (EIA) toxin A or B detection (culture test) with applying the EIA to stool samples alone (direct test).
Methods: In 2008, consecutive stool samples were cultured on C. difficile selective culture media: (1) medium I: Clostridium difficile-selective agar (CDSA; Becton Dickinson); (2) medium II: CLO agar (BioMérieux); (3) medium III: C. difficile agar according to Brazier (Oxoid). In addition, a direct test was performed (Ridascreen, r-Biopharm), which was also used to confirm toxin A or B production in the cultured C. difficile. New confidence bounds for sensitivities were applied, without assuming any perfect reference test or any conditional independence of the tests compared.
Results: Of 256 liquid stool samples, 18.4% were diagnosed as positive by at least 1 of the 4 tests; 12.8% were positive with culture medium I, 16.4% with medium II, and 13.6% with medium III, and 10.1% were positive by the direct test. Assuming culture tests to be at least as specific as the direct test yields an upper bound of 61% (upper 95% confidence bound (CB) 81%) for the sensitivity of the direct test. Assuming a prevalence of 15% yields sensitivity gains of the culture tests of at least 18% (lower 95% CB--4%) for medium I, 40% (lower 95% CB 21%) for medium II, and 23% (lower 95% CB 2%) for medium III.
Conclusions: Published high sensitivities of direct toxin A/B EIAs, up to 96%, and the correctness of the cytotoxicity test assumed for their estimation are doubtful. With culture medium II, sensitivity gains of at least about 20% are obtainable. Direct toxin A/B EIAs alone are insufficiently sensitive for the clinical diagnosis of C. difficile infections.