To Örebro University

This study investigates the performance of diagnostic methods for detection of Clostridioides difficile infection in Sweden, including impact of PCR ribotype on diagnostic performance. Between 2011 and 2016, a total of 17,878 stool samples from 26 laboratories were tested by either well-type enzyme immunoassays (EIAs), membrane bound EIAs, cell cytotoxicity neutralization assay (CTA), or nucleic acid amplification tests (NAATs) and subsequently cultured for C. difficile. Roughly half of the samples (9454/17878) were subjected to diagnostic testing both on the fecal sample and on the 1323 isolated C. difficile strains. All C. difficile isolates were typed by PCR ribotyping, and the isolates were classified as toxigenic or non-toxigenic based on the empirical knowledge of the association between toxin-positivity and ribotype. The overall sensitivity, specificity, and positive and negative predictive values were highest for NAATs and membrane EIAs. Ribotype-specific sensitivity varied greatly between methods and ribotypes. All methods had 100% sensitivity against ribotype 027 and 013. For other types, the sensitivity ranged from 33 to 85% in fecal samples and from 78 to 100% on isolates. For the most prevalent ribotypes (014, 020, and 001), the sensitivity varied between 38 and 100% in the fecal samples, with the lowest sensitivity observed for well-type EIAs and CTA. The large variation in diagnostic sensitivity implies that type distribution significantly affects the outcome when evaluating diagnostic performance. Furthermore, performing comparative studies of diagnostic tests in settings with high prevalence of ribotype 027 will overestimate the general performance of diagnostic tests.