A question about semantics really I guess. Say a clinical trial is conducted according to ITT principles and a Full Analysis Set is defined as (for example) subjects that take at least one dose of treatment and have one post-baseline measure. The sample size reported will be those subjects in each treatment group that have taken at least one dose of treatment and have one post-baseline measure, right? But am I correct in thinking that (unless you perform multiple imputation) this is really just an upper limit for your sample size, because there could be varying degrees of missingness on any of the post-baseline treatment effects you wish to estimate, which will in your statistical model automatically result in the dropping of those observations with missing data? e.g. Your FAS could be n = 100, but your 6 month treatment estimate may only be based on the data of 90 subjects (10 subjects dropped out before 6 months) and your 12 month treatment estimate may only be based on the data of 80 subjects (a further 10 subjects dropped out between 6 and 12 months).
You don't tend to see these model-specific sample sizes reported in the published results (unless I am missing something)
