Power analysis Q vs data analysis Q

Greetings.

I am trying to do a power analysis for a research proposal with multiple (say 5-15) assessments of clinical response over the treatment interval. There are two types of data in the literature from which to estimate effect sizes, both of which only look at baseline and end-of-treatment: Paired t-tests baseline to endpoint for active treatments, and 2-sample t-tests of difference scores between active and placebo treatments.

The analysis I had envisioned is a linear regression predicting clinical response from time of assessment, looking for an interaction between the linear trend line for active vs. placebo treatment. I'm having trouble figuring out how to use the effect sizes from the t-tests in the literature to get a plausible sample size proposal. Moreover, there are actually two active treatments (to be compared to the same placebo), but I'm presuming I'll do two separate tests of the interaction terms.

Can anyone help me think through this or can you point me somewhere?

Is there a better way to use all of the intermediate data, particularly if it is likely a swoop of some sort rather than a straight linear improvement?

Thanks for any thoughts,
Ron

PS: And if the the regression model contains covariates, how do I feed that into my power calculations?
.