10,000 is usually considered the number to start drawing formative ideas about long term profitability. I've seen discussions though where you are talking 100,000+ before conclusions can be drawn and I think this is reasonable. On the other hand VPIP and PFR start to take shape pretty quickly and there's other things that can be deduced from a smallish sample.
Tournaments are much more complicated to analyse using PT as they change structure as they proceed. Also you may find yourself shortstacked and pretty much fold 100 hands in a row, or have a decent stack early and play a lot of cheap hands. Different buy-ins, turbos, rebuys, qualifiers, SnGs all have different impact on your game and stats. This skews the data so more filters and more caution have to be used when analysing tourney hands in order to ensure the dataset you are using is consistent and valid. The data about amount won absolutely has to be in terms of BBs won rather than chips to make much sense.
Say you need 10,000 hands in comparable circumstances to draw conclusions you will need to play a lot more to have the data.
I play as tight as a gnats watsit early in SnGs, my VPIP is probably <10%. Once the first five levels are over it probably moves up to 35%+. To draw conclusions on the entirety of this data is really tricky so you need to filter carefully. I think PT is invaluable for any game but to be honest it has a far broader application in for ring players.
As a tourney player mainly I primarily use it with PAHUD to look at other people VPIP and PFR and occassionally scout for gaps in my own game.