Do all patients progress to Stage 5? If so, then you have two measures for each subject -elapsed time from entry to Stage 3 diagnosis, and elapsed time from Stage 3 diagnosis to Stage 5 diagnosis. For this situation, I would suggest a paired t test if you have no additional covariates, and a repeated measures analysis if there are additional covariates.
If the data is such that you don't have values in each interval for every patient, then a regular t test (possibly with unequal variances) if there are no additional covariates, and analysis of variance if there are additional covariates. It may be that you will need a mixed model approach if you need to account for clusters (e.g., hospitals or communities or such).
All of these can be done in SAS. It just depends on the design of the data and your question of interest.
SteveDenham.
