I tell my presenters to speak like no one can hear you....Because they can't.
It is said that a good engineer can get +10db of gain before feedback vs that of un-processed audio. I have never really bothered to try and find out how true that is, but I can say that I can usually find a way to get more gain than I need from a lav.
You have to use the NAG and PAG math. If your PAG is not higher than your NAG, you are fighting a losing battle. There are online calcs for this. Run some scenarios through them and see what you can get. And then if you are good, add another +10db to the PAG.
One of the tricks I use if I don't have a Neve 5045, is to use expansion. You set it up more like a gate, but you essentially set it so that it ducks the level down by a fixed amount when the person is not speaking and as soon as they start to make noise it opens back up. Using expansion is a little softer than a gate in how it opens and closes. This can usually let you get away with a few more db of fader level before feedback. It is a slippery slope though. If you are not on the ball, it will sail away on you in a hurry. Between expansion and EQ, I find I can go from just loud enough, to plenty loud. Another helpful trick is dynamic EQ, or multiband compression. A couple channels centered around the problem zone can allow you to have a natural sounding voice, while also ducking out the most problematic frequencies when speaking.