Prediction Tools for Unfavourable Outcomes in Clostridium difficile Infection: A Systematic Review

Valiquette L (2012) Prediction Tools for Unfavourable Outcomes in Clostridium difficile Infection: A Systematic Review. PLoS ONE 7(1): e30258. doi:10.1371/journal.pone.0030258 Prediction Tools for Unfavourable Outcomes in Clostridium difficile Infection: A Systematic Review Claire Nour Abou Chakra 0 1 Jacques Pepin 0 1 Louis Valiquette 0 1 Markus M. Heimesaat, Charite, Campus Benjamin Franklin, Germany 0 Competing Interests: Dr. Pepin has served on advisory boards for Pfizer , Wyeth, Ortho, Merck, Acambis , Iroko and The Medicines Company. Dr. Valiquette has served on advisory boards for Oryx, Iroko, Abbott and Wyeth, and has received compensation to conduct clinical trials involving antibacterials from Genzyme , Wyeth, Pfizer, BioCryst, Trius, Cempra , Optimer and Arpida. Claire Nour Abou Chakra has no competing interest to declare. This does not alter the authors' adherence to all the PLoS ONE policies on sharing data and materials 1 Department of Microbiology and Infectious Diseases, University of Sherbrooke , Quebec , Canada Context: Identifying patients at risk for adverse outcomes of Clostridium difficile infection (CDI), including recurrence and death, will become increasingly important as novel therapies emerge, which are more effective than traditional approaches but very expensive. Clinical prediction rules (CPRs) can improve the accuracy of medical decision-making. Several CPRs have been developed for CDI, but none has gained a widespread acceptance. Methods: We systematically reviewed studies describing the derivation or validation of CPRs for unfavourable outcomes of CDI, in medical databases (Medline, Embase, PubMed, Web of Science and Cochrane) and abstracts of conferences. Results: Of 2945 titles and abstracts screened, 13 studies on the derivation of a CPR were identified: two on recurrences, five on complications (including mortality), five on mortality alone and one on response to treatment. Two studies on the validation of different severity indices were also retrieved. Most CPRs were developed as secondary analyses using cohorts assembled for other purposes. CPRs presented several methodological limitations that could explain their limited use in clinical practice. Except for leukocytosis, albumin and age, there was much heterogeneity in the variables used, and most studies were limited by small sample sizes. Eight models used a retrospective design. Only four studies reported the incidence of the outcome of interest, even if this is essential to evaluate the potential usefulness of a model in other populations. Only five studies performed multivariate analyses to adjust for confounders. Conclusions: The lack of weighing variables, of validation, calibration and measures of reproducibility, the weak validities and performances when assessed, and the absence of sensitivity analyses, all led to suboptimal quality and debatable utility of those CPRs. Evidence-based tools developed through appropriate prospective cohorts would be more valuable for clinicians than empirically-developed CPRs. - In the decade that followed the emergence of the Clostridium difficile hypervirulent strain NAP1/BI/027 in North America and Western Europe, health professionals have been increasingly challenged by the burden of this infection, its frequent recurrences, severe complications and deaths [14]. Currently, the management of severe, complicated Clostridium difficile infection (CDI) is based on high-dose vancomycin, with or without intravenous metronidazole, intensive care unit (ICU) admission, vasopressor support and colectomy for a few selected patients [5,6]. Most patients present initially with similar symptoms, and identifying which ones might progress to these dreadful complications is difficult. After a long period of stagnation, novel therapeutic approaches are being developed for CDI. Fidaxomicin, recently licensed by the Food and Drug Administration, is more effective than vancomycin in avoiding recurrences [7,8]. Monoclonal antibodies were also proven to be effective in preventing recurrences, in a phase 2 trial [9]. Both fidaxomicin and monoclonal antibodies are unfortunately very expensive. Thus, it will become increasingly important to identify, early in the course of the disease, which patients would be most likely to benefit from these novel therapies, from closer follow-up, or both [10], ultimately to decrease CDIrelated morbidity and mortality. Clinical prediction rules (CPRs), which can improve the accuracy of medical decision-making, could address some of the aforementioned challenges in CDI management, and facilitate the conduct of clinical trials evaluating experimental therapeutic approaches. Several CPRs for CDI complications have been proposed over the years, but none has gained widespread clinical acceptance. We therefore performed a systematic review of all publications that aimed to derive or validate a CPR to predict recurrences, complications and mortality in patients diagnosed with CDI. Study selection A systematic review was performed according to PRISMA guidelines [11] (checklist S1) using an electronic search (Text S1) of all studies published since January 1978 (the year that C. difficile was identified as the etiological agent of pseudomembranous colitis [12,13]), in English, French or Spanish. The search was limited to humans and used the following online libraries and databases: Medline, PubMed, Cochrane, Embase and Web of Science. Furthermore, we reviewed abstracts submitted to conferences organised by the American Society for Microbiology, the Society for Healthcare Epidemiology of America, the Infectious Diseases Society of America, the Association of Medical Microbiology and Infectious Disease Canada, the Anaerobe Society of the Americas and the European Society of Clinical Microbiology and Infectious Diseases. In addition, the reference lists of identified CPRs were searched manually (crossreferencing). The final electronic search was performed on 30 October 2011. Inclusion criteria Publications from all sources were gathered in one file and duplicates were removed. A first screening of titles and abstracts followed by a full-text review were performed by CAC in order to identify studies that: i) focused on C. difficile as the main pathogen; ii) measured at least one relevant outcome: severity of the infection, complications, mortality, treatment failures or recurrences; and iii) developed or validated a model or risk score, a prediction rule, an index or a scale. Quality control on 10% of electronic search results was performed (LV) for the first screening of abstracts, as well as for all included studies. Reviewers had a good agreement concerning eligible studies (87%). Disagreements were resolved by a third party (JP). Data collection The following data were extracted by two reviewers (CAC and LV), from each included publication, into a standardized matrix: definitions of main outcomes, description of the study (...truncated)