I've used GPower3.1 and Optimal Design for calculating a priori sample sizes for studies, but this design is more complex and I can't seem to find the appropriate category in either GPower or Optimal Design. Does anyone know a good resource for calculating needed sample size for more complex designs?

Here are the specs: alpha = .05 two-tailed, Beta = .20, variance .10, expected effect size .25. Design is a 4-arm trial with 3 tx conditions and 1 control condition. Two states with several clinic sites will first determine condition severity (mild vs. moderate) then randomly assign individual patients to 1 of 4 tx conditions within each severity category. We will assess each individual 4 times (1 week, 1 month, 3 months, 6 months) and compare omnibus effect of txs and do pair-wise comparisons between 4 tx group means (6 comparisons).

How many patients do I need in each tx arm? Within each state or total? Does anyone know how to characterize this design in Optimal Design? Seems to be mostly geared to cluster randomized trials, and this is intervention at the individual level. Repeated measures does not capture the characteristic of the 4 tx arms. Or do I just use an alpha adjustment to control family-wise error rate (.05/6)?