You can use whatever information you collected at the baseline to test whether the drop-outs are significantly different from those who remain in the study.
If you find that your drop-outs are "missing at random" (i.e., no systematic differences) then you have less to worry about.
I fully agree with David's comments. In my opinion, it is also important to prespecify in the protocol (and/or the Statistical Analysis Plan) how dropout patients will be handled in the statistical analysis. The EMA guideline "Guideline on Missing Data in Confirmatory Clinical Trials" can provide you some hints about it (http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2010/09/WC500096793.pdf)
Dear Shakil,
In general I am joining the opinion of colleagues and would like to add just the following:
Drop-outs is a serious problem for any randomized trial. First of all these patients are to be included into intent-to-treat analysis but at the same time these patients if they have not received minimal "complete" treatment (if such minimal complete treatment or minimal duration of treatment, or minimal cumulative dose of the treatment, was defined in the study protocol) are not included into per-protocol analysis and thus they would/could compromise statistical power of the trial. The situation is even worse if, for example the primary endpoint of the study is overall survival - in this case the date of the last contact with the patient is entered into the database as the date of death, or if the endpoint is the Progression Free Survival - the date of the last contact is considered the date of progression of main disease. Due to that for any trial the sponsor should consider/estimate/predict definite drop-out rate and increase from the very beginning the sample size by its per cent to avoid having at the end of the trial study results, which are not powered enough or even not statistically significant.
Regards,
Maxim Belotserkovskiy, MD
The following paper provides a comprehensive insight on the types of missing data and the analytical approaches to deal with it. I hope it would be of some interest to you.
Dziura JD, Post LA, Zhao Q, Fu Z, Peduzzi P. Strategies for dealing with missing data in clinical trials: from design to analysis. The Yale journal of biology and medicine. 2013 Sep;86(3):343.
I suggest increasing the sample size with an estimated number of dropouts. Would it be helpful to avoid applying statistical methods?
- Badges
- Science topic
- Similar topics
- Correction