I just got out of a Fiber Optics engineering class last semester so lets see how much I remember XD
When it comes to fiber optics of a short length (like this), dispersion doesnt matter any longer. Dispersion is when the light wavelengths traveling down the fiber take different paths (think how a smaller sine wave travels less distance than a larger sine wave) and so travel different distances and the signal on the output is literally stretched lengthwise from the original signal. This is a major problem for analog signals. Digital signals do have some problem from it but it only becomes a factor when it comes to how fast the signal can be sent down the fiber before overlap of 0 and 1 occurs. Because the signal being sent in audio is very low compared to that of most fibers bandwidth limitation, this shouldnt be a factor.
Attenuation is the next major factor. Just like sound, attenuation is the diminished amplitude of the signal being sent. This only becomes a problem at long distances (longer than what you would be using fiber for in audio). If the distance becomes to great then a repeater station must be set into place so that the signal can be reamplified and retransmitted. This is more of a problem for analog than digital.
Im not going to thoroughly into these because theres distinction for these when it comes to multimoded and singlemoded fibers (meaning only one mode of light can propagate, or many). I will say though that multimoded fibers arent as good as single moded fibers because the bandwidth is smaller, though the cost is greater. Singlemode>GRIN Multimode>Multimode (GRIN is Graded Index where the refractive values of the fiber core are graded so as to lessen dispersion).
There is the factor of back scattering, but that shouldnt be a major factor for something like this.
If its a digital signal being sent a distance less than what would be used for a WAN network, then a plastic/plastic multimode fiber should be more than adequate. Just like digital TV vs analog TV, you have the wall effect of the signal for digital. Either the signal will be "loud and clear", or it will be completely nonsensical.
...though this is Head-Fi, so if you really REALLY want the best of the best, you would go for a dispersion shifted, vapor formed glass/glass single mode fiber with a LASER diode source at 1550nm with a Faraday isolater in the line.
Hope this cleared things up =X I can clarify anything if there is confusion.