If you really want to compare means, you are in the world of parametric statistics. Most of the non-parametric tests examine medians or other characteristics of the distribution.
Friedman's 2-way analysis of variance can handle some types of repeated measures without assuming normality. I'm not sure whether or where it is in SAS.
If you have enough data, you could convert the outcomes to ranks and rely on the central limit theorem to support conclusions based on a parametric analysis.
You might also be able to use a variance stabilizing transformation and apply the parametric methods.