If your choice of terminology is accurate, then you generated a Geodatabase Replica to ship to the regional office. The replica generation process includes a named version which ends up receiving the regional edits once returned.
Importantly, there are specific requirements to use the replica in a disconnected state. (Be sure to review the geodatabase replication help topic in detail.) Specifically, you maychoose to use replicas which are registered to existing data...then GlobalIDs are required in both target GDbs to ensure correct feature registration.
Here's what happens next: the regional office makes the editor changes, then they export replica changes (xml messages) which are transmitted, then synchronized back at the main office GDb...these changes should end up in the version which originated the replica. Once the XML replica synchronization changes are processed into the source replica version, normal GDb reconcile/verification/post processing can occur. (Please also review the GDb Replica Synchronization help section here.)
Note that, just as you asked, only the regional office edits are transmitted and scnced back to the base which makes the network throughputmore efficient. There are web-based methods and APIs to move the replica data back and forth efficiently.
I have frequent success practicing the basic replication above in a single GDb instance with differing Owners...thats similar to working between base and regional GDbs but skips thenetwork bandwidth. Also, there are tools in ESRIs forum notes to review the replica XML if diagnostics are needed.