Generally speaking, here are the steps.
1. Query logs on ISA server
2. Determine the URLs that users used to access RSS feeds
3. Analyse the URLs and work out the patterns to match most of the URLs to be blocked
4. Define one or more URL filters using the patterns
5. Apply the filters
6. Monitor the traffic to see if the filters work
7. Refine the filters if necessary.
However, I just checked the source code of my Google Reader page, and found it uses "feed/
http://" to locate the RSS feeds. I am not sure if the string would appear in each GET request to remote HTTP servers as I don't have a traffic monitor installed at the moment. You may try test (query) this on your ISA server.
Regards,