This is a very broad question, which might be easier to reverse. What problems do you have in your setup? Is the load too high? Are there continuous disk operations?
In general, using NRPE results in the Icinga masters to execve(2) the plugin for each check. Depending on your check intervals, this may result in lots of unnecessary load. Using the Icinga 2 agent would reduce this, since it sends network commands to the hosts on its own. However, this would be a huge change and if it currently works for you, I would not touch it.
On a similar note, Icinga DB tires to address some IO issues of IDO. Thus, switching might help. Especially for Icinga DB, we have started an operations section, also containing performance related information.
But if everything works in its current state, I would keep it this way