I'll answer my own question. We ended up using SNMP and walking through each channel on the circuits which was processed via a Perl script to add them up based on each circuit. In the cases of multiple circuits, we used a shell script to sum each circuit together to get an overall trunk group number. We then fed those numbers to MRTG to graph it out. So far it's working well except for the fact that we only get a picture at that moment in time so we could theoretically miss a huge spike in between polls. Oh well, sometimes you don't get to eat the cake.
