Leaving out the theoretcials, the first thing you have to realize is that you may be talking apples and oranges. What's important for low light and noise is not the size of the pixel site, but the actual size of the light sensitive sensor...
see this article for a brief explanation....
What is a backlit CMOS sensor? | News | TechRadar
So the first question becomes which camera has the largest photo receptor, that is different from which has the biggest photo sites as determined by how much room each occupies on the chip.
I've also seen a breakdown of a Sony technology that makes the photosensitive part of the sensor almost the whole size of the space allotted by placing the circuitry under the sensitive array.
That just means you need a lot more knowledge than I'd know where to find before you even start.
But just for a comparison, if you measure resolution by line widths per picture height (lw/ph)
A D610 FF camera with 24 MP will produce ~2,800 lw/ph from a raw image under ideal conditions based on a pixel def. of 6016x4016. So the D600 produces 2800/6016 or .465 distinct lines per pixel.
A D7100 also 24 Mp but APS-c will also produce 2800 pixels from raw.. from 6000x4000...
Nikon D7100 Review - Exposure
A Pentax K-3 resolved about 2700 lw/ph..
Pentax K-3 Review: Now Shooting! - Exposure
So in terms of lw/ph, there is no advantage to a larger sensor. At least looking at current models, all tested with the same lens as far as I know.
When you start discussing artifacts like moire and noise at low light, the smaller sensor seems to be at a disadvantage.
A Zeiss Makro-Planar T* 100mm f/2 ZF (ZE) (on Canon EOS) records 3776 lw/ph on a 21 Mp camera.
The Sigma 70 macro tested at ƒ4 also tested out at 3736 on the Canon body, and that is the lens used to test the D610, D7100 and K-3 tests posted above... there may be other attributes that make the lens desirable...
I don't know what lens as use to test the 645D but it also resolves down to 3700... but the conclusion would that a Zeiss Makro-Planar T* 100mm f/2 ZF (ZE) (on Canon EOS 21MP), is as good as an image taken on a 40 Mp 645D... pretty scary.
A D800 on the other hand also probably with a Sigma 70 produces 3400 lw/ph
Nikon D800 Review - Exposure, this is also similar to an A7r.
Looking at all the numbers, they just don't add up in any real way, except that as pointed out above, a top quality lens may be the biggest contributing factor in maximum resolution...
My conclusion would be , test the lens you want on the camera you want and see what it gives you... there are just way to many variables to make sense of, trying to come up with some kind of mathematical formula to understand it all.