Lycoming, in their 'Lycoming Flyer' recommend engine failures in twins be simulated using mixture.
They state it causes much less stress on the engine by still allowing air into the cylinders to provide a more normal pressurised environment in the cylinders. This gives a different loading against the piston/conrod/bearing/mating surfaces than closing the throttle from high power settings.
I take slight issue to Hudson's comment re "...no record of the close shaves that inevitably occur...". If there is no record of these incidents, then it is only an assumption that they occurred.
The counterpoint to that argument is to say that there is no record of all the safe times that this method was used. It is only when the rate of incidents is different that one can say one method is, or is not, more risky.
I suspect that there have been incidents from closing the mixture - but I also suspect that there have been incidents from closing the throttle. Without data to compare the rate then it is only supposition to state one that one or other is more, or less, dangerous.
Personally, I suspect I've had more trouble with restoring power quickly after a period simulating an engine failure due to fouled plugs at idle than when using mixture. This has varied from a/c type to a/c type & also with temp/altitude etc. In one type it was necessary to retard the mixture to restore smooth running while the throttle was at idle. On the other hand in another it was a cow to start after being shutdown.
I would like to see a properly researched trial between the two methods. I doubt that it will happen though.