• Jennifer S Gewandter, Michael P McDermott, Andrew McKeown, Kim Hoang, Katarzyna Iwan, Sarah Kralovic, Daniel Rothstein, Ian Gilron, Nathaniel P Katz, Srinivasa N Raja, Stephen Senn, Shannon M Smith, Dennis C Turk, and Robert H Dworkin.
• aDepartment of Anesthesiology University of Rochester, Rochester, NY, USA bDepartment of Biostatistics and Computational Biology, University of Rochester, Rochester, NY, USA cCity University of New York-Hunter College, New York City, NY, USA dDepartment of Pain Medicine, MD Anderson Cancer Center, Houston, TX, USA eDepartment of Anesthesiology, Queen's University, Kingston, ON, Canada fAnalgesic Solutions, Natick, MA, USA gTufts University, Boston, MA, USA hDepartment of Anesthesiology and Critical Care Medicine, Johns Hopkins University, Baltimore, MD, USA iLuxembourg Institute of Health, Strassen, Luxembourg jDepartment of Anesthesiology and Pain Medicine, University of Washington, Seattle, WA, USA.
• Pain. 2016 Nov 1; 157 (11): 2544-2551.

AbstractCross-over trials are typically more efficient than parallel group trials in that the sample size required to yield a desired power is substantially smaller. It is important, however, to consider some issues specific to cross-over trials when designing and reporting them, and when evaluating the published results of such trials. This systematic review evaluated the quality of reporting and its evolution over time in articles of cross-over clinical trials of pharmacologic treatments for chronic pain published between 1993 and 2013. Seventy-six (61%) articles reported a within-subject primary analysis, or if no primary analysis was identified, reported at least 1 within-subject analysis, which is required to achieve the gain in power associated with the cross-over design. For 39 (31%) articles, it was unclear whether analyses conducted were within-subject or between-group. Only 36 (29%) articles reported a method to accommodate missing data (eg, last observation carried forward, n = 29), and of those, just 14 included subjects in the analysis who provided data from only 1 period. Of the articles that identified a within-subject primary analysis, 21 (51%) provided sufficient information for the results to be included in a meta-analysis (ie, estimates of the within-subject treatment effect and variability). These results and others presented in this article demonstrate deficiencies in reporting of cross-over trials for analgesic treatments. Clearer reporting in future trials could improve readers' ability to critically evaluate the results, use these data in meta-analyses, and plan future trials. Recommendations for proper reporting of cross-over trials that apply to any condition are provided.

20 keywords...

hide…