Such a thing can't actually be stated without a controlled study. And of course, such a study could never be performed.
Guessing if you randomly sampled 200k people of all ages 18-50 let's say, races, etc., so long as they were generally considered healthy and did the following study:
- half randomly assigned to be handcuffed behind back, laid chest down on a hard surface, and cop the size of Chauvin kneels on their shoulder blade for 9 mins
- other half randomly gets the same, except kneels on neck for 9 mins
you'd see the neck group with vastly larger death rate. What would that rate be? That is the question. I would not be surprised if at least 20% died.