At the moment there is a single requirement from InterConnect, assuming that Study Level Meta Analyses will be made available in a release September, which is the ability to run Cox models on each dataset separately and meta-analyse the results.
The reason for doing this is that people are uncomfortable when they hear about the piecewise Poisson alternative (even though I know it almost always gives the same results). But it also removes the need to run the Lexis function and to set the bucket sizes.
When speaking to Paul I know there is then an assumption about the baseline hazard then being the same across all studies - but is that only a problem when trying to fit a single model across all studies?