Objective: In randomised controlled trials of weight loss interventions, high dropout rates are a problem resulting in a large amount of missing outcome data. It is common for participants with missing data to be excluded from analysis (complete-case analysis). We aim to demonstrate how published complete-case results can be used to explore how study results would change depending on assumptions about dropout weight loss.
Design and Methods: Our methods are based on three extensions to a method for obtaining baseline observation carried forward (BOCF) results from complete-case results. Our first extension is a generalisation to any dropout weight loss. Second, we show that it is not necessary to assume that dropout weight loss is the same in each treatment arm. Third, we show that variation in dropout weight loss can be incorporated. Using these extensions, sensitivity analyses to the missing data can be conducted via the use of plots.
Results: We demonstrate our methods using two examples of published results from studies of weight loss interventions. We also demonstrate how the BOCF method could be useful to meta-analysts.
Conclusion: Byusing simple plots, readers can explore how different assumptions about dropout weight loss affect the results of published weight loss trials.
Read More Here