Effect size for a trial with pre and post measurements

Question

I have recently been told two different ways to prepare data for calculating effect size (Hedge's g) of a placebo-controlled trial:

1. use post-intervention means of the assessment values for both groups (treatment and control)
2. use the means difference between the assessment values at baseline and the post-intervention assessment for both groups (treatment and control)
3. use the pre and post means for treatment without regard for control

Which is correct? Please give authoritative source(s) if you can.

Answer 1

Sample size computation for complex trial designs is not a bread and butter issue, as many factors must be considered, on top of, of course, the chosen alpha and beta (https://www.crcpress.com/Sample-Sizes-for-Clinical-Trials/Julious/p/book/9781584887393).

Indeed, focusing on controlled trials with repeated measures, several key assumptions on baseline effects, post-treatment effects, time-wise interactions, and treatment-wise interactions may be explicitly considered.

There are several computing approaches which can prove useful, including some R packages (e.g. https://www.r-bloggers.com/power-and-sample-size-for-repeated-measures-anova-with-r/).

My recommendation is to have good preliminary data in order to be really informed on baseline effects, time-wise changes, and intervention-wise effects. Then move on with calculations.

Otherwise, simply reducing all assumptions to a comparison between delta in control group vs. delta in experimental group (where delta is the difference between post-treatment and baseline effect in the chosen group) may lead to workable and understable computations.