As I believe it lag is generated by a couple of things, first is a models polycount.
Higher poly models generate higher lag. (though realisticly this shouldn't cause huge problems, unless of course you have say a unit with a polycount of 25, and you replace it with one with 100, but you're still creating 200 units of each) I don't know the exact numbers when it comes to polys that generate harsh lag because I'm not a modeler, but overall I don't think its a huge issue you really need to worry about.
Another thing that generates lag is triggers. The more triggers running, being run, or being proced (however you want to say it) the more lag. Especially if the triggers calls other triggers and are rather heavy and long. Heroes without spells wont really damage this all that much, but if you think about it when a hero levels, hardcode is running to change its level, so heroes generate a very small amount of extra usage. However, this issue probably wont effect lag at all, even with 200 units.
As far as handling 200 handles go, i've asked the question myself but with a higher number, and was given the answer that people have been able to handle very large numbers of "static" units without any harsh or even moderate lag at all. Those numbers being in the 4 to 5 hundreds.
Another possibility (though I think this might be more of a self-generated myth) is that heroes generate more lag to the user because the minimap puts more strain on the users processors with all the blinky glows. This is easily changeable though, if it even affects (you'll have to ask someone else if you really think it matters).
In conclusion, I would say that Yes, using heroes instead of units does generate "more" lag, but No, that lag is not nearly excessive enough to cause any problems.