Mostly LR is used for server side performance test.
Can LR been used for the end user performance testing for a web based application? That is to measuring how long does it take for each page fully loaded after a user action: a click for instance.
This link will provide some useful information http://stackoverflow.com/questions/5567634/load-runner-and-browser-client-performance check this out.
So QTP does do the tests inclusive of client side processing time. And when executed through LR will still be using object based interactions and hence is very scalable as compared to options below.
But the flipside is that it requires a machine per vuser (unless you are using VMs / Terminal Services based Load Generators) PLUS it requires QTP licenses on top of the LR protocol license.
Given the above, you may also consider Remote Desktop Procol as an alternative. Flip side on RDP is that the scripts will be co-ordinate based and hence prone to sync issues on replay.
Other option is Citrix based access of the application, where the objects are slightly more context driven than RDP. Flipside is that Citrix needs to be licensed.
I think for that is LR the wrong tool.