How to use published complete case results from weight loss studies in a missing data sensitivity analysis

Objective: In randomized 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). The aim of this article is to demonstrate how published complete-case results can be used to explore how study results would change depending on assumptions about dropout weight loss.

Methods: The methods are based on three extensions to a method for obtaining baseline observation carried forward (BOCF) results from complete-case results. The first extension is a generalization to any dropout weight loss. Second, to show that it is not necessary to assume that dropout weight loss is the same in each treatment arm. Third, to 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: The methods are demonstrated using two examples of published results from studies of weight loss interventions. It is also shown how the BOCF method could be useful to meta-analysts.

Conclusion: By using simple plots, readers can explore how different assumptions about dropout weight loss affect the results of published weight loss trials.