What you really need is an experiment where you run many engines to TBO, some with LOP procedures, some with the old ROP procedures. The conditions need to be as close as possible to identical for each sample i.e. no significant difference in pilot population etc. You then compare the number of engine problems in each sample.
American Airlines had 400,000,000 of data collected and ran TBO's several times longer LOP than they or the military could achieve ROP. Deakin San is probably one of the few old enough to remember. (sorry JD
)
If anyone had proof of which you seek they did.