R help - Aug 2011 - Multiple endpoint (possibly group sequential) sample size calculation

Hello everyone,

I need to do a sample size calculation. The study two arms and two endpoints.
The two arms are two different cancer drugs and the two endpoints reflect
efficacy (based on progression free survival) and toxicity.

Until now, I have been trying to understand this in terms of a one-arm design,
where the acceptable rate of efficacy might be 0.40, the unacceptable rate of
efficacy might be 0.20, the acceptable rate of non-toxicity might be 0.85, and
the unacceptable rate of non-toxicity might be 0.65. Then one would pick an
alpha for the probability of accepting a poor response, an alpha for the
probability of accepting a toxic drug, and a beta for the probability of
rejecting a good drug.

I'm not really sure how that sort of thinking translates into a two-arm
design though.

Ideally, I'd like the calculation to be based on a group sequential design
with two stages, but that's certainly not necessary, and I'd be very
happy to learn how to do things both with and without this extra element.

Any help with this would be greatly appreciated.

Thanks,

Paul

On Aug 4, 2011, at 2:03 PM, Paul Miller wrote:
> 
> Hello everyone,
> 
> I need to do a sample size calculation. The study two arms and two
endpoints. The two arms are two different cancer drugs and the two endpoints
reflect efficacy (based on progression free survival) and toxicity.
> 
> Until now, I have been trying to understand this in terms of a one-arm
design, where the acceptable rate of efficacy might be 0.40, the unacceptable
rate of efficacy might be 0.20, the acceptable rate of non-toxicity might be
0.85, and the unacceptable rate of non-toxicity might be 0.65. Then one would
pick an alpha for the probability of accepting a poor response, an alpha for the
probability of accepting a toxic drug, and a beta for the probability of
rejecting a good drug.
> 
> I'm not really sure how that sort of thinking translates into a two-arm
design though.
> 
> Ideally, I'd like the calculation to be based on a group sequential
design with two stages, but that's certainly not necessary, and I'd be
very happy to learn how to do things both with and without this extra element.
> 
> Any help with this would be greatly appreciated.
> 
> Thanks,
> 
> Paul
Hi Paul,

I am not clear that there is a current R package that will handle both of these
considerations, though will stand to be corrected if wrong.

A Google search using "sample size calculation multiple endpoints"
yields some possible theoretical papers that might be helpful, such as:

  http://www.ncbi.nlm.nih.gov/pubmed/20687162
  http://www.ncbi.nlm.nih.gov/pubmed/17674404

I have used the gsDesign package on CRAN to assist with group sequential designs
with a single primary endpoint. There was just a review in this month's The
American Statistician which covered several software implementations, including
gsDesign:

  http://pubs.amstat.org/doi/abs/10.1198/tast.2011.10213

There is also a CRAN Task View that might be helpful:

  http://cran.r-project.org/web/views/ClinicalTrials.html

My various books on group sequential designs are at home, so am unable to check
them at the moment, but pretty sure that at least Jennison and Turnbull (1999)
have a chapter on multiple endpoints.

HTH,

Marc Schwartz