Presented at the CTN Design & Analysis Workshop, CTN Steering Committee Meeting, Bethesda, MD, March 15, 2011
Neal Oden, PhD (Data and Statistics Center, EMMES), Gaurav Sharma, PhD (Data and Statistics Center, EMMES), Paul C. VanVeldhuisen, PhD (Data and Statistics Center, EMMES), Paul G. Wakim, PhD (Center for the Clinical Trials Network, NIDA).
This presentation from the "CTN Design & Analysis" workshop at the 2011 Steering Committee Meeting addresses the problem of missing data from CTN trials. The focus is mostly on primary outcomes data, which may be missing for a variety of reasons, including discontinuation of the study, outcomes undefined for some participants (such as quality of life measures after death), or attrition. Though CTN studies are focused on efficacy, not perfection (that is, it's not "Does treatment work if perfectly delivered?" but instead "Is this a good treatment strategy?"), researchers should still strive to collect complete data from all participants, even those who do not complete the study, as results will never be believable, no matter how sophisticated the statistical method, if there is too much missing data. A variety of approaches for dealing with missing data are discussed, including ways to design trials to help minimize the likelihood of missing data. Ways to analyze missing data are also provided, including repeated-measure designs, linear and quadratic time trend or spline models, and the importance of sensitivity analysis. The presentation uses protocol CTN-0010 to provide a case study about ways to work with and around missing data. (Presentation, PPT, English, 2011)
Keywords: Clinical trials - Methods | Data collection | Missing data | Research design | Statistical analysis | CTN Steering Committee meeting, March 2011
Document No: 665
Submitted by Paul G. Wakim, NIDA CCTN, 3/24/2011.