Bitrate means nothing by itself. Different encoders will have different results at the same bitrate. The most advanced/transparent at any given rate today is usually the latest version Lame, using the tested encoding presets.
In reality, 256kbps encoding with Lame will result in most music being transparent to most people during most of the time. It's not safe to say always, because you can always find a 'codec buster' recording, though they are very rare. For example, a harpsichord recording, with not much reverb and none or very little accompaniment instruments, will tend to be much more difficult to encode to MP3 as compared to most anything else. There are a handful of instruments/sounds that will be difficult to encode.
One famous DBT involved original studio master vs. 256kbps Lame, using B&W 801 monitors in a studio, using many experienced audio engineers as the test subjects. The end result was no one was able to achieve a score better than 'chance guessing'; around 50%.
I have tried to DBT randomly picked music samples that I encoded with Lame at 256kbps, and was not able to do better than chance guessing under blinded conditions. This is while using extremely high resolution playback hardware. I did not use known difficult to encode samples - which would of course result in a different outcome. My purpose was to determine if I could discern audible differences using the actual music that I enjoy.
In any event, I do prefer to use lossless music files when practical. It would seem the only real cases where MP3s are needed is for portable audio devices and for general distribution on the internet. Hard drive space is so cheap now that it is practical to store most audio collections in lossless formats.
-Chris