I suppose your question is between NSCA and NRPE. It depends on how you want to setup your distributed environment, one is an "active" solution, the master server says "do this check" and waits for the results, in the second system it's the remote server doing the check and sending the result to the main server.
I've had a look at icinga recently and from their presentation it looks like the system interfaces have been simplified a lot (in regards to DB and other internal messaging), could be worth a look on a big system if you have the chance to make some tests, the config files should be quite identical yet.
I have no idea about Centreon, but the past experience (a couple of years back so YMMV) wasn't positive with any kind of frontend to nagios, be it in the administrative part or in the visualization. The config stuff mostly broke the files in mysterious and unexpected ways while there was little benefit to the visualization, the few i found interesting were nagvis (layering images to the maps to have a more immediate look for what is wrong) and one with a name similar to nagiosgraph (the one to graphically create the nagios maps).