Let's say that top speed for a human is 10m/s. (Faster than me, but Carl Lewis can do it.)
The centripetal acceleration is [tex]\frac{v^2}{r}[/tex]. In order for the runner not to fall from the top of the loop he must be at least in free fall at the top of the loop. So we get
[tex]\frac{100\frac{m^2}{s^2}}{r} > 9.81 \frac{m}{s^2} \rightarrow r < \frac{100}{9.81} m[/tex]
So if the human could maintain top speed all the way up, then it would be possible for someone like Carl Lewis to run a loop that was 10m high.
This is, of course, unrealistic. According to the same calculations, someone who can run 5m/s should be able to run a 3m high loop. (This is not possible -- at least not for normal humans)
There are major factors that the calculation ignores - one is that the amount of energy that goes into going up the loop while maintaining that speed is quite large -- certainly larger than what a human can produce -- that there will be a loss of friction as the runner goes up the loop which increases the problems with energy consumed by climbing , and also that, for a circular loop, the runner would be pulling 2G's while running at the bottom of the loop.
Depending on how you define running a loop, running up a wall, and then flipping back would qualify. This is something that people can certainly do (watch "Singing in the rain" for a spectacular example.) Based on footage like that I would say that at least some humans can run loops.