Statistical errors in microleakage studies in operative dentistry. A survey of the literature 2001-2009

The aim of this study was to assess the literature on microleakage of direct restorations in operative dentistry indexed in the ISI Web of Science, in order to assess the robustness of the statistical methodology used. Our database included 226 scientific papers (published between 2001 and 2009 in 22 journals) from the journal citation report categories 'Dentistry, Oral Surgery and Medicine' and 'Materials Science, Biomaterials'. We reviewed all articles to find potential mistakes that are commonly made at different stages in the scientific research process. Microleakage was assessed quantitatively in 50 (22.2%) studies and qualitatively in 176 (77.8%) studies. In all studies reviewed the statistical methods used were appropriate for the category attributed to the outcome variable, but in 13% of the total, the chi-square test or parametric methods were inappropriately used afterwards. When the appropriate statistical methods were applied in studies that had originally employed inappropriate methods to analyse their data, and in which the authors provided raw data, an alteration of the conclusions was necessary in 15.4% of these re-analysed studies. This survey also showed that the statistical methodology applied varies considerably for similar experimental designs. This could have an effect on statistical results; hence, a more standardized methodology should be implemented.